Whether the Chrome OS Accessibility Menu is set to be shown regardless of the state of a11y features.(checked once 45 secs after startup).

Whether the Chrome OS Autoclick feature is on (checked once 45 secs after startup).

If the user has enabled Autoclick, this is the delay set by the user for autoclicks to occur, in milliseconds.

Whether the Chrome OS High Contrast mode feature is on (checked once 45 secs after startup).

Whether the Chrome OS Large Cursor feature is on (checked once 45 secs after startup).

Whether the Chrome OS Screen Magnifier feature is on (checked once 45 secs after startup).

Whether the Chrome OS Spoken Feedback feature is on (checked once 45 secs after startup).

Whether the Chrome OS Virtual Keyboard feature is on (checked once 45 secs after startup).

Whether Windows system settings show that high-contrast mode is enabled and the user has selected a light-on-dark color scheme (checked once 45 secs after startup). This causes Chrome to prompt the user with a bubble to optionally install a High Contrast extension and theme.

Whether Chrome has enabled accessibility support because the user passed the --force-renderer-accessibility flag on the command-line (checked once 45 secs after startup).

Whether Chrome has enabled accessibility support because it detects supported assistive technology running, or due to being manually enabled via a command-line flag (checked once 45 secs after startup).

Whether Windows system settings show that audio descriptions are enabled (checked once 45 secs after startup).

Whether the third-party JAWS screen reader is running (checked once 45 secs after startup).

Whether the third-party NVDA screen reader is running (checked once 45 secs after startup).

Whether the third-party System Access To Go screen reader is running (checked once 45 secs after startup).

Whether Windows system settings show that a screen reader is running (checked once 45 secs after startup). Note that this does not necessarily mean that Chrome has detected a supported screen reader and has enabled its accessibility mode.

Whether Windows system settings show that Sticky Keys are enabled.

Whether the third-party ZoomText screen magnifier is running.

The number of times v2 packaged apps are launched grouped by extension_misc::AppLaunchBuckets. See also Extensions.AppLaunch.

Interactions with the App Launcher promo dialog.

The number of searches that are started in the app list. This is gathered each time the app list search box transitions from empty to non-empty.

The type of app list search result that was opened by the user. This is gathered per click of a search result.

Number of active touch-points when a new touch-point is added.

The show type of the active window tracked over time by logging on a regular basis (30 minutes).

User-initiated action taken that affects docked windows such as docking, undocking, minimizing, restoring, closing or just dragging a docked window.

Source (mouse, touch or unknown) of the user-initiated action for docked windows.

Number of all docked windows or panels including hidden or minimized. Recorded on every user action that interacts with docked windows.

Number of large (wider than dock maximum width) windows that had to be shrunk to get docked among the visible docked windows. Recorded on every user action that interacts with docked windows.

Number of docked visible panels. Recorded on every user action that interacts with docked windows.

Number of visible docked windows or panels. Recorded on every user action that interacts with docked windows.

Time elapsed between instances of docking, undocking or any other action affecting docked state of a window.

Width of the docked area in pixels. Recorded every time it changes after a user window resize operation is completed.

The gesture-events recognized and dispatched by the browser gesture recognizer.

The gesture-events recognized and dispatched by the browser gesture recognizer for various UI components.

The type of the window which is put into immersive fullscreen. Immersive fullscreen is entered via the F4 key.

The current state of the shelf (alignment) tracked over time by logging on a regular basis (30 minutes), this is used instead of log in or shelf usage to track users that do not lock/unlock or log in frequently and use a small number of browser instances or otherwise infrequently interact with the shelf launcher.

The current state of the shelf (alignment) when the shelf launcher is used to launch an app/window/etc, this is used instead of log in to give data on users that do not lock/unlock or log in frequently.

Deprecated 12/2013 in r239809, and replaced by Ash.TouchDuration2.

The duration of a touch-sequence.

The interval between touch-move events.

The distance between touch-move events.

The position of the touch-events along the X axis.

The position of the touch-events along the Y axis.

The radius of a touch event.

The interval between the end of a touch-sequence and the start of the next touch-sequence.

The number of rapid touch-starts that happened within a short interval. Logged once for each such burst group.

The amount of time the Alt key is held after pressing Alt+Tab to begin cycling through windows.

The number of items (single windows or groups of windows such as panels) in the window selection. Window selection occurs when a user begins alt-tabbing or presses the overview button (F5 key).

The amount of time between uses of window selection to switch between windows. Window selection is entered by alt-tabbing or by pressing the overview button (F5 key).

The amount of time spent in overview mode. Overview mode is engaged when lingering on a window while alt-tabbing or by pressing the overview button. The time is measured from the moment the windows begin animating to a thumbnail size preview to when a window is selected or selection is canceled.

Count of DnsAttempts before DnsTransaction completes with failure.

Count of DnsAttempts before DnsTransaction completes successfully.

Whether DnsConfigService::OnConfigChange actually corresponded to a change in DnsConfig.

Duration of time between calls to DnsConfigService::InvalidateConfig.

Duration of time spent parsing DnsConfig.

Counts of results of parsing DnsConfig in DnsConfigServicePosix.

Whether DnsConfig was parsed successfully.

Counts of results of parsing DnsConfig in DnsConfigServiceWin.

Whether the first valid DnsConfig included a rogue nameserver.

Counts of specific error codes returned by DnsTask if a subsequent ProcTask succeeded, at the end of a streak of failures after which the DnsClient was disabled.

TRUE counts the events when a valid DnsConfig is received and used to enable DnsClient, while FALSE counts the events when DnsClient is disabled after a series of successful fallbacks from DnsTask to ProcTask.

Duration of time spent by ProcTask in failing fallback resolutions.

Duration of time spent by ProcTask in successful fallback resolutions.

Whether there was a valid DNS configuration at the start of a job which eventually completed successfully.

Whether DnsHosts were parsed successfully.

Whether DnsConfigService::OnHostsChange actually corresponded to a change in DnsHosts.

Duration of time between calls to DnsConfigService::InvalidateHosts.

Duration of time spent parsing DnsHosts.

Counts of results of parsing DnsHosts in DnsConfigServiceWin.

The size of the HOSTS file observed before each attempt to parse it.

Time elapsed between the time the HostResolverImpl::Job was created and the time the Job was started (using DnsClient).

Time elapsed between the time the HostResolverImpl::Job was created and the time the Job was started (using DnsClient). Includes only Jobs which had priority HIGHEST when started.

Time elapsed between the time the HostResolverImpl::Job was created and the time the Job was started (using DnsClient). Includes only Jobs which had priority IDLE when started.

Time elapsed between the time the HostResolverImpl::Job was created and the time the Job was started (using DnsClient). Includes only Jobs which had priority LOW when started.

Time elapsed between the time the HostResolverImpl::Job was created and the time the Job was started (using DnsClient). Includes only Jobs which had priority LOWEST when started.

Time elapsed between the time the HostResolverImpl::Job was created and the time the Job was started (using DnsClient). Includes only Jobs which had priority MEDIUM when started.

Time elapsed between the last time the priority of a HostResolverImpl::Job changed (when a Request was attached or detached) and the time the Job was started (using DnsClient).

Time elapsed between the last time the priority of a HostResolverImpl::Job changed (when a Request was attached or detached) and the time the Job was started (using DnsClient). Includes only Jobs which had priority HIGHEST when started.

Counts of results of parsing addresses out of DNS responses in successful DnsTransactions.

Counts of specific error codes returned by DnsTask if a subsequent ProcTask succeeded.

Duration of time taken by DnsTask in resolutions that failed. Excludes time spent in the subsequent fallback.

Counts of the overall results of using asynchronous DNS in HostResolverImpl. This only includes jobs started with valid DNS configuration and excludes synchronous resolutions (as IP literals, from cache, and from HOSTS).

Duration of time taken by DnsTask in resolutions that succeeded.

Same as AsyncDNS.ResolveSuccess, but limited to pure IPv4 lookups.

Same as AsyncDNS.ResolveSuccess, but limited to pure IPv6 lookups.

Same as AsyncDNS.ResolveSuccess, but limited to IPv4/IPv6 lookups.

Count of servers in DnsConfig. Recorded on every new DnsSession, which is created on DNS change.

Index in DnsConfig of the failing server, recorded at the time of failure.

Count of server failures after network change before first success in the DnsSession. Recorded at the time of first success.

Count of server failures after success until the end of the session. Server has reported success at some point during the session. Recorded at the end of the DnsSession.

Count of server failures before success. This is NOT the first success in the DnsSession. Recorded at the time of success.

Count of server failures without success until the end of the session. Server has never reported success during the DnsSession. Recorded at the end of the DnsSession.

The current server is "good" and does not have to be skipped.

Duration of time taken in failing calls to AddressSorter in dual-stack resolutions using DnsTask.

Duration of time taken in successful calls to AddressSorter in dual-stack resolutions using DnsTask.

The number of names from the search name list consumed during a successful transaction (QTYPE A only).

The number of names left on the search name list at the end of a successful transaction (QTYPE A only).

The number of names on the search name list at the start of a transaction (QTYPE A only).

Duration of time taken by DnsTCPAttempt in failed attempts. Excludes timeouts.

Duration of time taken by DnsTCPAttempt in successful attempts.

Difference between RTT and timeout calculated using Histogram algorithm.

Difference between timeout calculated using Histogram algorithm and RTT.

Difference between RTT and timeout calculated using Jacobson algorithm.

Difference between timeout calculated using Jacobson algorithm and RTT.

Duration of time that would be spent waiting for lost request using Histogram algorithm.

Duration of time that would be spent waiting for lost request using Jacobson algorithm.

Duration of time since a HostResolverImpl::Resolve request to the time a result is posted. Excludes canceled, evicted, and aborted requests. Includes cache hits (recorded as 0). Excludes speculative requests.

Duration of time taken in failing DnsTransactions. This includes server failures, timeouts and NXDOMAIN results.

Duration of time taken in successful DnsTransactions. This includes all NOERROR answers, even if they indicate the name has no addresses or they cannot be parsed.

Same as AsyncDNS.TransactionSuccess but limited to A query type.

Same as AsyncDNS.TransactionSuccess but limited to AAAA query type.

TTL of the resolved addresses, as in the response received from the server. For results served from local cache, the TTL is from the original response.

Duration of time taken by DnsUDPAttempt in failed attempts. Excludes timeouts.

Duration of time taken by DnsUDPAttempt in successful attempts. Includes responses arriving after timeout, if multiple attempts are allowed.

Duration of time since the last empty config result to the time a non-change OnConfigChange is received.

Duration of time since the last empty config result to the time a non-change OnHostsChange is received.

The result of DnsConfigService watch. Counts STARTED on every initialization and FAILED_* on any failure.

Whether the browser compositor uses GPU or the software renderer.

Deprecated as of 8/2013.

Measures the frequency of user interactions with the Autocheckout bubble, which prompts users to invoke Autocheckout on supported websites.

Deprecated as of 8/2013.

Measures the frequency of final states reached in Autocheckout buy flow.

Deprecated as of 8/2013.

The state of the Autocheckout dialog when it was dismissed.

Deprecated as of 8/2013.

Measures the time elapsed between when the user submitted the Autocheckout dialog and when the Autocheckout flow, or filling process, concluded.

Deprecated as of 8/2013.

Measures the time elapsed between when the user submitted the Autocheckout dialog and when the Autocheckout flow concluded, in cases where the flow failed.

Deprecated as of 8/2013.

Measures the time elapsed between when the user submitted the Autocheckout dialog and when the Autocheckout flow concluded, in cases where the flow succeeded.

Deprecated as of 8/2013.

The initial state of a user that's interacting with a freshly shown Autocheckout dialog.

Deprecated as of 8/2013.

User interactions with the Autofill popup shown while filling an Autocheckout dialog.

Deprecated as of 8/2013.

Measures the frequency of security warnings and errors in the Autocheckout dialog.

Deprecated as of 8/2013.

Measures the duration for which an Autocheckout dialog was shown.

Deprecated as of 8/2013.

Measures the duration for which an Autocheckout dialog was shown, in cases where the user ended up canceling out of the dialog.

Deprecated as of 8/2013.

Measures the duration for which an Autocheckout dialog was shown, in cases where the user ended up accepting the dialog.

Deprecated as of 8/2013.

Measures how users are interacting with the Autocheckout dialog UI.

Deprecated as of 8/2013.

Measures the duration of time it takes for the Autocheckout UI to be actionable by the user after it is shown.

Deprecated as of 8/2013.

Measures the frequency of errors in communicating with the Google Online Wallet server.

Deprecated as of 8/2013.

Measures the frequency of required user actions returned by the Google Online Wallet server.

Deprecated as of 8/2013.

Measures time taken to download the Autocheckout whitelist file.

Deprecated as of 8/2013.

Measures time taken to download the Autocheckout whitelist file in case the download was failed.

Deprecated as of 8/2013.

Measures time taken to download the Autocheckout whitelist file in case the download was succeeded.

The number of address suggestions shown in the Autofill popup.

Deprecated as of 3/2011, replaced by Autofill.CreditCardInfoBar.

The Autofill credit card info bar was accepted.

Deprecated as of 3/2011, replaced by Autofill.CreditCardInfoBar.

The Autofill credit card info bar was denied.

The relative frequency with which users accept, deny, or ignore the Autofill credit card info bar prompt.

Measures the adoption of the HTML autocomplete type hint specification (see http://is.gd/whatwg_autocomplete for more details). For each fillable form detected, logs whether that form includes author-specified type hints.

Time elapsed between the user's first interaction with a form and the form's submission, for an autofilled form.

Time elapsed between the user's first interaction with a form and the form's submission, for a non-autofilled form.

Time elapsed between form load and form submission, for an autofilled form.

Time elapsed between form load and form submission, for a non-autofilled form.

Tracks whether Autofill is enabled on page load for a page containing forms.

Tracks whether Autofill is enabled when Chrome launches.

Deprecated as of 3/2011, replaced by Autofill.StoredProfileCount.

The number of Autofill address profiles a user has.

Deprecated as of 3/2011, replaced by Autofill.Quality.

The quality of the AutoFill implementation.

The quality of the Autofill implementation.

The quality of Autofill's heuristic field type detection.

The quality of Autofill's heuristic field type detection, broken down by the specific field type. Fields with multiple possible types (based on the stored Autofill data) are logged as having ambiguous type.

The overall quality of the Autofill field type predictions.

The overall quality of the Autofill field type predictions, broken down by the specific field type. Fields with multiple possible types (based on the stored Autofill data) are logged as having ambiguous type.

The quality of the Autofill server's field type detection.

The quality of the Autofill server's field type detection, broken down by the specific field type. Fields with multiple possible types (based on the stored Autofill data) are logged as having ambiguous type.

TBD.

Deprecated as of 6/2011, replaced by Autofill.ServerExperimentId.Query.

The experiment ID received in response to an Autofill server query.

The experiment ID received at the time of an Autofill upload.

Deprecated as of 3/2011, replaced by Autofill.ServerQueryResponse.

The usefulness of AutoFill server information.

The usefulness of Autofill server information.

The number of Autofill addresses a user has stored, measured at launch time.

Measures the frequency of various events in the Autofill user interaction flow. By comparing frequencies, we can compute several interesting "user happiness" metrics.

Counts the number of simulataneously connected Bluetooth devices. Used to direct testing efforts, and by our UI team to determine appropriate UI sizes.

Records the method used to pair each Bluetooth Device. Used to direct our testing efforts.

Records the result of pairing each Bluetooth Device. Used to understand whether we are having significant problems with Bluetooth pairing and seeing errors more commonly than we should.

Logs the depth of the bookmark in the bookmark tree hiearchy every time a bookmark is launched. Depth indicates how many levels below a permanent bookmark folder the bookmark was found in (e.g. a bookmark immediately in the bookmark bar has depth 1).

Records the context type names used to create canvas rendering contexts.

Records the number of times the cast button was shown to the user. The value will be true if the button is enabled, and false if the button is disabled. Note that depending on the current UX, it's possible that we hide the button entirely if it's disabled, so it's possible for the false values to be 0.

Records the number of times the cast button was shown to the user when the video is fullscreened. The value will only be recorded on entering fullscreen. The value will be true if the button is enabled, and false if the button is disabled. Note that depending on the current UX,it's possible that we hide the button entirely if it's disabled, so it's possible for the false values to be 0.

Records the result of a request to play remotely. The value will be true if the playback succeeded, and false if there was an error.

Records the percentage of the video left at the time the remote playback is stopped. This will be recorded when the playback is stopped by the user, or when it's stopped by the cast device.

The count of cellular device activation failures (Chrome OS).

The count of cellular device activation tries (Chrome OS).

The count of cellular reconnect failures during activation (Chrome OS).

The count of cellular device reconnect tries during activation (Chrome OS).

The count of successful cellular plan established (Chrome OS).

The count of initiated cellular device setup starts (Chrome OS).

The count of failed cellular plan setup tries (Chrome OS).

The count of failed cellular plan purchases (Chrome OS).

The count of successfully completed cellular plan purchases (Chrome OS).

Deprecated as of 8/2013. This histogram only considered the leaf certificate expiry date as a proxy for whether a certificate was in-scope for the BRs, but did not consider the issuance date. As some CAs have issued long-lived certs prior to the BRs, this disproportionately reported those certs as being subject to the BRs, but non-compliant, when in reality they're not subject.

Information about the certificate algorithms and sizes in use on the web, to examine compliance with the CA/Browser Forum requirements and security best practice.

Information about the certificate algorithms and sizes in use on the web, to examine compliance with the CA/Browser Forum requirements and security best practice. This histogram considers the notBefore as the issuance date, for purposes of what requirements apply.

Indicates the execution phase the browser was in when the browser crashed.

Deprecated as of 11/2013.

Indicates the execution phase the browser was in when browser didn't exit cleanly.

The total number of times the browser process has attempted to generate a crash dump. This should be the sum of Chrome.BrowserDumpsWithCrash and Chrome.BrowserDumpsWithNoCrash.

The number of times the browser process has attempted to generate a crash dump because of an actual browser crash.

The number of times the browser process has attempted to generate a crash dump in a non-crashing (i.e., reporting only) context.

Deprecated 8/2013. No longer tracked.

The default search engine selected by a user not in the search engine dialog experiment.

Deprecated 8/2013. No longer tracked.

The default search engine selected by a user in the search engine dialog experiment.

Deprecated 8/2013. No longer tracked.

The default search engine selected by a user in slot 1 of a randomized search engine dialog.

Deprecated 8/2013. No longer tracked.

The default search engine selected by a user in slot 2 of a randomized search engine dialog.

Deprecated 8/2013. No longer tracked.

The default search engine selected by a user in slot 3 of a randomized search engine dialog.

Deprecated 8/2013. No longer tracked.

The default search engine selected by a user in slot 4 of a randomized search engine dialog.

The actions to enable or disable services sending synced notifications. Synced Notification Sending services can be individually disabled by the user in the Chrome Notification center settings dialog.

The length of time between a dangerous download appearing on the downloads shelf, and the "Discard" button being clicked.

The length of time between a dangerous download appearing on the downloads shelf, and the "Dismiss" button being clicked.

The length of time between the external protocol dialog being shown and the "Launch Application" button being clicked.

The length of time between a download appearing on the download shelf, and the user opening it by clicking the item or pressing return.

Time between "Report and Discard" button being shown and it being clicked.

The length of time between a dangerous download appearing on the download shelf, and the "Keep" button being clicked.

Deprecated as of 4/2013, experiment confirmed correctness of our patch.

Counts how often the user writes or reads from the clipboard and whether the write was from an incognito window or not.

Connectivity Diagnostics App: WiFi signal strength recorded during NIC_SIGNAL_STRENGTH test.

The "Strength" property of a WiFi signal is a partially-reversible function that linearly maps the RSSI range -120dBm to -20dBm to Strength values from 0 to 100.

HTTP latency seen by the Connectivity Diagnostics.

HTTP latency is computed using the chrome.socket API to make an HTTP GET request to the /generate_204 page of three randomly generated Google hostnames (*-ccd-testing-v4.metric.gstatic.com). The time taken from issuing the HTTP request to receiving a response is clocked in JavaScript and the arithmetic mean of the three times is used as the HTTP latency.

Resolution latency seen by the Connectivity Diagnostics.

Resolver latency is computed by using the chrome.dns API to query three randomly generated Google hostnames (*-ccd-testing-v4.metric.gstatic.com). The random hostnames guarantees that there will be no caching of DNS hostnames. The time taken from issuing the DNS request to receiving a response is clocked in JavaScript and the arithmetic mean of the three times is used as the resolver latency.

Connectivity Diagnostics App: Outcome of the connectivity tests.

Connectivity Diagnostics App: Amount of time taken to run each of the connectivity tests.

Deprecated as of 9/2013. Experiment to measure control characters in cookies is finished.

When parsing a cookie, indicates if control characters were present in any of the cookie values and if any of the cookie values were invalid. Specifically, checks that all of the parsed values are valid according to the valid token definition in Section 2.2 of RFC2616 which specifies a token must have no separators (i.e. no characters from the following string, ignoring the starting and ending single quote: '()<>@,;:\"/[]?={} \t') and no control characters.

The duration in seconds between a cookie getting evicted (due to the number of cookies exceeding a domain limit), and subsequently reinstated.

Deprecated as of 9/2013. Experiment to measure control characters in cookies is finished.

Indicates whether a cookie attribute pair was set with both a valid key and a valid attribute value or not. For the key, this implies that it was a valid token as defined in Section 2.2 of RFC2616 which specifies a token must have no separators (i.e. no characters from the following string, ignoring the starting and ending single quote: '()<>@,;:\"/[]?={} \t') and no control characters. For the value, this implies that it contained no control characters and no semicolon.

Deprecated as of 9/2013. Experiment to measure control characters in cookies is finished.

Indicates whether a cookie name was set with a valid token. A valid token is defined in Section 2.2 of RFC2616 which specifies a token must have no separators (i.e. no characters from the following string, ignoring the starting and ending single quote: '()<>@,;:\"/[]?={} \t') and no control characters.

Deprecated as of 9/2013. Experiment to measure control characters in cookies is finished.

Indicates whether a cookie value was valid or invalid when there was an attempt to set it, where a valid value is defined in RFC 6265 as ASCII characters excluding controls, whitspace, comma, semicolon, and backslash.

Deprecated as of 12/2013. Default pinned apps trial is finished.

Chrome OS shelf clicks.

The type of archive file that Chrome OS cros-disks daemon is requested to mount.

The media type of removable device that Chrome OS cros-disks daemon is requested to mount.

The type of file system that Chrome OS cros-disks daemon is requested to mount.

Records the number of times when first-run dialog was shown.

Index of furthest step that was reached during tutorial. Since order of steps could change eventially and new steps could apear we use index here instead of step name.

The total time that user spent on first-run tutorial.

The time that user spent on some step of tutorial.

Tracks the way how user left tutorial.

Records the number of times when first-run tutorial has been launched.

Counts various events that trigger Chrome to bypass the fallback configuration of the data reduction proxy.

Counts various events that trigger Chrome to bypass the primary configuration of the data reduction proxy.

Counts various outcomes of requesting the data reduction proxy's probe URL.

Samples which method was used by the user to dismiss the proxy promo. This is sampled when the promo leaves view, with the sampled value depending on which of four possible controls the user used.

Samples of user interactions with the ON/OFF switch in the settings menu for reducing data usage. Only the setting changes between entering the reducing data usage setting menu and leaving the menu will be sampled. So if a user enters the menu with OFF and leaves it with OFF, it is counted as one OFF to OFF conversion regardless of how many times he or she toggles the ON/OFF switch.

Samples of the state of the data reduction proxy on Chrome startup. The proxy will either be unavailable (the feature hasn't been rolled out to this user yet), not enabled (the feature is available but the user doesn't have it turned on), or enabled (the feature is enabled and turned on).

TBD - Not run automatically yet, so this is just a placeholder for future metrics collection. Any samples collected here represent users running diagnostics manually.

Shows the success and failure rates of the DiskSpace recovery step that runs on recovery startups. The recovery step attempts to guarantee the DiskSpace test, which checks that the disk space in the volume where the user data directory normally lives is not dangerously low, would pass on the next startup.

TBD - Not run automatically yet, so this is just a placeholder for future metrics collection. Any samples collected here represent users running diagnostics manually.

Shows the success and failure rates of the JSONBookmarks recovery step that runs on recovery startups. The recovery step attempts to guarantee the JSONBookmarks test, which makes sure that the JSON-encoded Bookmarks file is properly formed, would pass on the next startup.

Shows the success and failure rates of the JSONLocalState recovery step that runs on recovery startups. The recovery step attempts to guarantee the JSONLocalState test, which makes sure that the JSON-encoded Local State file is properly formed, would pass on the next startup.

Shows the success and failure rates of the JSONPreferences recovery step that runs on recovery startups. The recovery step attempts to guarantee the JSONPreferences test, which makes sure that the JSON-encoded Preferences file is properly formed, would pass on the next startup.

TBD - Not run automatically yet, so this is just a placeholder for future metrics collection. Any samples collected here represent users running diagnostics manually.

Shows the success and failure rates of the PathDictionaries recovery step that runs on recovery startups. The recovery step attempts to guarantee the PathDictionaries test, which makes sure that the path to the Dictionaries directory exists and has the right permissions, would pass on the next startup.

Shows the success and failure rates of the PathLocalState recovery step that runs on recovery startups. The recovery step attempts to guarantee the PathLocalState test, which makes sure that the path to the Local State file exists and has the right permissions, would pass on the next startup.

Shows the success and failure rates of the PathResources recovery step that runs on recovery startups. The recovery step attempts to guarantee the PathResources test, which makes sure that the path to the Resources directory exists and has the right permissions, would pass on the next startup.

Shows the success and failure rates of the PathUserData recovery step that runs on recovery startups. The recovery step attempts to guarantee the PathUserData test, which makes sure that the path to the User Data directory exists and has the right permissions, would pass on the next startup.

Shows the success and failure rates of the SQLiteIntegrityAppCache recovery step that runs on recovery startups. The recovery step attempts to guarantee the SQLiteIntegrityAppCache test, which checks the integrity of the App Cache database, would pass on the next startup.

Shows the success and failure rates of the SQLiteIntegrityArchivedHistory recovery step that runs on recovery startups. The recovery step attempts to guarantee the SQLiteIntegrityArchivedHistory test, which checks the integrity of the Archived History database, would pass on the next startup.

Shows the success and failure rates of the SQLiteIntegrityCookie recovery step that runs on recovery startups. The recovery step attempts to guarantee the SQLiteIntegrityCookie test, which checks the integrity of the Cookie database, would pass on the next startup.

Shows the success and failure rates of the SQLiteIntegrityDatabaseTracker recovery step that runs on recovery startups. The recovery step attempts to guarantee the SQLiteIntegrityDatabaseTracker test, which checks the integrity of the Database Tracker database, would pass on the next startup.

Shows the success and failure rates of the SQLiteIntegrityHistory recovery step that runs on recovery startups. The recovery step attempts to guarantee the SQLiteIntegrityHistory test, which checks the integrity of the History database, would pass on the next startup.

Shows the success and failure rates of the SQLiteIntegrityNSSCert recovery step that runs on recovery startups. The recovery step attempts to guarantee the SQLiteIntegrityNSSCert test, which checks the integrity of the NSS Certificate database, would pass on the next startup.

Shows the success and failure rates of the SQLiteIntegrityNSSKey recovery step that runs on recovery startups. The recovery step attempts to guarantee the SQLiteIntegrityNSSKey test, which checks the integrity of the NSS Key database, would pass on the next startup.

Shows the success and failure rates of the SQLiteIntegrityThumbnails recovery step that runs on recovery startups. The recovery step attempts to guarantee the SQLiteIntegrityThumbnails test, which checks the integrity of the Thumbnails database, would pass on the next startup.

Shows the success and failure rates of the SQLiteIntegrityWebData recovery step that runs on recovery startups. The recovery step attempts to guarantee the SQLiteIntegrityWebData test, which checks the integrity of the Web Data database, would pass on the next startup.

TBD - Not run automatically yet, so this is just a placeholder for future metrics collection. Any samples collected here represent users running diagnostics manually.

Count of the number of times diagnostics recovery is invoked or not, and how it was invoked. A sample is added to this histogram once for each startup of Chrome.

TBD - Not run automatically yet, so this is just a placeholder for future metrics collection. Any samples collected here represent users running diagnostics manually.

Shows the success and failure rates of diagnostics for the DiskSpace test that runs on recovery startups. The DiskSpace test checks that the disk space in the volume where the user data directory normally lives is not dangerously low.

TBD - Not run automatically yet, so this is just a placeholder for future metrics collection. Any samples collected here represent users running diagnostics manually.

Shows the success and failure rates of diagnostics for the JSONBookmarks test that runs on recovery startups. The JSONBookmarks test checks to make sure that the JSON encoded bookmarks file is properly formed.

Shows the success and failure rates of diagnostics for the JSONLocalState test that runs on recovery startups. The JSONLocalState test checks to make sure that the JSON encoded Local State file is properly formed.

Shows the success and failure rates of diagnostics for the JSONPreferences test that runs on recovery startups. The JSONPreferences test checks to make sure that the Preferences file is properly formed.

TBD - Not run automatically yet, so this is just a placeholder for future metrics collection. Any samples collected here represent users running diagnostics manually.

Shows the success and failure rates of diagnostics for the PathDictionaries test that runs on recovery startups. The PathDictionaries test checks makes sure that the path to the Dictionaries folder exists and has the right permissions.

Shows the success and failure rates of diagnostics for the PathLocalState test that runs on recovery startups. The PathLocalState test checks makes sure that the path to the Local State folder exists and has the right permissions.

Shows the success and failure rates of diagnostics for the PathResources test that runs on recovery startups. The PathResources test checks makes sure that the path to the Resources folder exists and has the right permissions.

Shows the success and failure rates of diagnostics for the PathUserData test that runs on recovery startups. The PathUserData test checks makes sure that the path to the User Data folder exists and has the right permissions.

Shows the success and failure rates of the SQLiteIntegrityAppCache test that runs on recovery startups. The test checks the integrity of the App Cache database.

Shows the success and failure rates of the SQLiteIntegrityArchivedHistory test that runs on recovery startups. The test checks the integrity of the Archived History database.

Shows the success and failure rates of the SQLiteIntegrityCookie test that runs on recovery startups. The test checks the integrity of the Cookie database.

Shows the success and failure rates of the SQLiteIntegrityDatabaseTracker test that runs on recovery startups. The test checks the integrity of the Database Tracker database.

Shows the success and failure rates of the SQLiteIntegrityHistory test that runs on recovery startups. The test checks the integrity of the History database.

Shows the success and failure rates of the SQLiteIntegrityNSSCert test that runs on recovery startups. The test checks the integrity of the NSS Certificate database.

Shows the success and failure rates of the SQLiteIntegrityNSSKey test that runs on recovery startups. The test checks the integrity of the NSS Key database.

Shows the success and failure rates of the SQLiteIntegrityThumbnails test that runs on recovery startups. The test checks the integrity of the Thumbnails database.

Shows the success and failure rates of the SQLiteIntegrityWebData test that runs on recovery startups. The test checks the integrity of the Web Data database.

TBD - Not run automatically yet, so this is just a placeholder for future metrics collection. Any samples collected here represent users running diagnostics manually.

Histogram comparing the various types of diagnostic test failures when diagnostic tests are run. Note that some types of test failures cause the rest of the tests to be skipped.

Whether an extension has been wiped out.

How many external extensions get wiped out as a result of the Sideload Wipeout one-time initiative.

Whether any extension got wiped out as a result of the Sideload Wipeout one-time initiative.

The user selection in the Sideload Wipeout bubble, grouped by the UmaWipeoutHistogramOptions enum.

The age of the cache's files (wall time).

The age of the cache's files (wall time). Media-specific cache.

The age of the cache's files (wall time). AppCache.

The age of the cache's files (wall time). ShaderCache.

Deprecated.

The total time it takes to perform a payload IO operation, for the regular disk cache.

The attempt which completed after the job was already cancelled.

The attempt which completed after the job was already cancelled OR the attempt that has finished after host resolution was already completed by an earlier attempt.

Duration of time taken in OS resolutions for actual navigations. These attempts which completed after the job was already canceled OR after the job was already completed by an earlier attempt. Note that cached resolutions may provide low (0ms?) resolution times.

The attempt that has not resolved the host successfully.

The attempt that resolved the host first and the resolution was not successful.

The attempt that resolved the host first and the resolution was successful.

The attempt that has resolved the host successfully.

Duration of time taken in OS resolutions that succeeded and were requested for actual navigations. These attempts which completed after the job was already canceled OR after the job was already completed by an earlier attempt. Note that cached resolutions may provide low (0ms?) resolution times.

This histogram shows the time saved by having spawned an extra attempt, when the first attempt didn't finish before retry attempt.

The time left to expiration of an entry when it is removed while compacting the HostCache.

The time since expiration of an entry when it is removed while compacting the HostCache.

The time since expiration of an entry when it is removed on lookup.

Error status when an empty address list was found in OnLookupComplete().

When either a pre-resolution was not done recently enough to provide benefit, or the the corresponding pre-resolution is still pending, this histogram shows the duration of time used to resolve a hostname as not existing during a failed attempt to navigate to (GET) a URL. In newer versions, if the hostname has never been found as a link during a page scan, and it has a referring URL, then it is added to referrer list data structure (hoping we'll do better next time).

When either a pre-resolution was not done recently enough to provide benefit, or the the corresponding pre-resolution is still pending, this histogram shows the duration of the duration of time used to resolve a hostname to navigate to (GET) a URL. In newer versions, if the hostname has never been found as a link during a page scan, and it has a referring URL, then it is added to referrer list data structure (hoping we'll do better next time).

Time elapsed between the time the HostResolverImpl::Job was created and the time the Job was started (a getaddrinfo call was dispatched to the thread pool).

Time elapsed between the time the HostResolverImpl::Job was created and the time the Job was started (a getaddrinfo call was dispatched to the thread pool). Includes only Jobs which had priority HIGHEST when started.

Time elapsed between the last time the priority of a HostResolverImpl::Job changed (when a Request was attached or detached) and the time the Job was started (a getaddrinfo call was dispatched to the thread pool).

The duration of time used (most recently) to pre-resolve a hostname, when the prefetched resolution was apparently evicted from the cache. The included samples only list pre-resolution times when the later navigations/fetches took in excess of 15ms.

Replaced by DNS.PrefetchFoundNameL.

Deprecated 2/2010, and replaced by DNS.PrefetchResolution

The duration of time used by the DNS pre-resolving threads to resolve a host name via the network. Any resolutions that are faster than 15ms are considered to be local cache hits, not requiring network access, and are not included in this histogram. This histogram is most useful for estimating the typical cost of a name resolution, but it also estimates the total number of network-based resolutions induced by this feature. Not all these resolutions prove helpful (i.e., the user does not always actually visit the resolved hostnames).

Replaced by DNS.PrefetchNegativeHitL.

The duration of time saved due to DNS pre-resolving in the "name not found" case. Time "savings" shown in the histogram are defined to be the difference between the DNS pre-resolution duration, and the DNS resolution duration seen during a navigation. These cache hits only list events where the DNS pre-resolve duration for a host was in excess of 15ms (i.e., the network was consulted), and the actual DNS resolution (when a user attempted to navigate to a link with the same host name) took less than 15ms (i.e., the network was not consulted), which means the gain was a result of a "cache hit" in the OS cache. For some users with LANs, all negative results (even when the DNS cache might otherwise help) take about 2.5 seconds (due to timeouts for netbios broadcasts), and hence no savings are possible (or shown) for such users in this category.

Replaced by DNS.PrefetchPositiveHitL.

The duration of time saved due to DNS pre-resolving in the "name was found" case, and induced by either a page scan for a link or an omnibox entry by the user. Time "savings" shown in the histogram are defined to be the difference between the DNS pre-resolution duration, and the DNS resolution duration seen during a navigation. These cache hits only list events where the DNS pre-resolve duration for a host was in excess of 15ms (i.e., the network was consulted), and the actual DNS resolution (when a user attempted to navigate to a link with the same host name) took less than 15ms (i.e., the network was not consulted), which means the gain was a result of a "cache hit" in the OS cache.

The duration of time spent by a proposed resolution waiting in the queue to be resolved. This number is in addition to any DNS resolution time that may come later.

The duration of time saved due to DNS pre-resolving in the "name was found" case, and induced by predicting (using referrer lists) that a resolution was needed. Time "savings" shown in the histogram are defined to be the difference between the DNS pre-resolution duration, and the DNS resolution duration seen during a navigation. These cache hits only list events where the DNS pre-resolve duration for a host was in excess of 15ms (i.e., the network was consulted), and the actual DNS resolution (when a user attempted to navigate to a link with the same host name) took less than 15ms (i.e., the network was not consulted), which means the gain was a result of a "cache hit" in the OS cache.

When, due to congestion avoidance, a queued pre-resolution is abandoned (recycled) without actually being resolved, this histograms records the age in the queue of that entry. Only times over 2 seconds are recorded in this histogram.

Counts of successes and failures of OS resolutions in various categories.

Duration of time taken in OS resolutions for actual navigations. Note that cached OS resolutions may provide low (0ms?) resolution times.

Same as DNS.ResolveFail, but limited to pure IPv4 lookups.

Same as DNS.ResolveFail, but limited to pure IPv6 lookups.

Same as DNS.ResolveFail, but limited to IPv4/IPv6 lookups.

Duration of time taken in speculative OS resolutions. Note that cached OS resolutions may provide low (0ms?) resolution times.

Duration of time taken in speculative OS resolution that succeeded. Note that cached resolutions may provide low (0ms?) resolution times.

Duration of time taken in OS resolutions that succeeded and were requested for actual navigations. Note that cached resolutions may provide low (0ms?) resolution times.

Same as DNS.ResolveSuccess, but limited to pure IPv4 lookups.

Same as DNS.ResolveSuccess, but limited to pure IPv6 lookups.

Same as DNS.ResolveSuccess, but limited to IPv4/IPv6 lookups.

Deprecated as of 5/2013.

Counts of hits and misses in the DNS cache and DNS jobs pool of wasted HostResolverImpl::Jobs that could be avoided by always resolving using AF_UNSPEC.

In some cases, such as when content arrives with embedded references to other servers, the prefetch system can't (or doesn't) attempt to pre-resolve the hostnames. As an example, a visit to www.cnn.com will fetch content with references to about 12 additional hostnames, none of which are currently anticipated. Such resolutions are termed "Unexpected Resolutions," and the durations associated with those DNS resolutions are shown below. Future features may attempt to learn (from prior experience locally, or from server provided hints), what secondary hostname resolutions should be done when a primary resolution (or navigation) takes place. This histogram shows what the potential savings are that "remain on the table" until we employ some of these more advanced features.

In some cases, such as when content arrives with embedded references to other servers, or when a page (such as one in SSL) preclude scanning and prefetching, the prefetch system can't (or doesn't) attempt to pre-resolve the hostnames. As an example, a visit to www.cnn.com will fetch content with references to about 12 additional hostnames, none of which might be anticipated. Similarly, clicking on a link in an SSL page won't be anticipated (since scanning in not allowed by default). Such resolutions are termed "Unexpected Resolutions," and the durations associated with those navigation induced DNS resolutions are shown below. If a referring URL is available for the navigation, the relationship to the referring URL was recorded, and future navigations to the referring hostname would have induced a pre-resolution of hostname that caused an entry below. Such any entry may facilitate future listing in the ReferredPositiveHit histogram.

Status of DNS probe updates sent to a DNS error page.

Renamed 7/2013 to DnsProbe.ProbeDuration.

Time between starting and finishing DNS probe.

Removed 7/2013.

Time between starting and finishing DNS probe when NCN says we're offline.

Removed 7/2013.

Result of DNS probes sent by the probe service when NCN says we're offline.

Removed 7/2013.

Time between starting and finishing DNS probe when NCN says we're online.

Removed 7/2013.

Result of DNS probes sent by the probe service when NCN says we're online.

Renamed 7/2013 to DnsProbe.ProbeResult. (Also switched to the full DnsProbe.ProbeStatus enum.)

Result of DNS probes sent by the probe service.

Removed 7/2013.

Elapsed time of DNS probes that return PROBE_BAD_CONFIG.

Removed 7/2013.

Whether the only nameserver in the system DNS config was 127.0.0.1 when the probe result was BAD_CONFIG.

Removed 7/2013.

The result of the system probe job when the overall probe result was BAD_CONFIG.

Removed 7/2013.

The number of nameservers in the system DNS config when the probe result was BAD_CONFIG.

Removed 7/2013.

Elapsed time of DNS probes that return PROBE_NO_INTERNET.

Removed 7/2013.

Elapsed time of DNS probes that return PROBE_NXDOMAIN.

Removed 7/2013.

Elapsed time of DNS probes that return PROBE_UNKNOWN.

Time between starting and finishing DNS probe.

Result of DNS probes sent by the probe service.

Number of certs loaded from domain bound cert database.

Time spent loading domain bound cert database.

The size, on disk, of the domain bound cert database as it is being loaded.

Time spent generating a domain bound cert.

Combined time for GetDomainBoundCert retrieval (both synchronous and asynchronous).

Time for asynchronous retrieval (from the GetDomainBoundCert call until completion callback is called).

Time for synchronous GetDomainBoundCert cert retrieval.

Result of GetDomainBoundCert function.

Whether the domain-bound certs sqlite database was killed succesfully when an unrecoverable error was detected.

Counts of SSL client sockets broken down by support for Domain Bound Certificates TLS extension. Counts only connections with full handshakes, resumed sessions are not counted.

Longest time spent by requests waiting for load of domain bound cert database.

Number of requests that waited for load of domain bound cert database.

The length of downloads for serves that accept byte ranges.

The length of downloads for serves that do not specify whether the accept ranges, or have invalid ranges specified.

The length of downloads for serves that do not accept ranges.

The actual bandwidth (per read) of a download.

Downloads extension API function calls.

Disk bandwidth (defined as total bytes divided by the amount of time blocked on write or close on the file descriptor) seen for a single download.

Overall bandwidth seen for the download. Note that this is measured at the point at which the file is written, and so will not take into account the time costs of activities that occur after file write is completed (e.g. safe browsing scanning).

The percentage of the potential bandwidth actually used (per read) of a download. An entry of 100% implies that Chrome was the limiting factor in download speed.

The number of downloads in history at the time it is cleared.

Content-Disposition header features. The presence of a Content-Disposition header, use of 'name', 'filename' and 'filename*' parameters, and string encoding schemes are counted for each unthrottled download. The total number downloads is Download.Counts[5] (Initiated and Unthrottled).

Types of images that are downloaded.

Content types that are downloaded.

Various individual counts in the download system; see DownloadCountType for details.

Various individual counts in the download system, for example the number of downloads blocked by throttling from the DownloadRequestLimiter.

User chose to save a download which was marked dangerous. Grouped by the type of danger.

User chose to save a download which was marked DANGEROUS_FILE. Grouped by the type of file.

A download which was marked DANGEROUS_FILE was discarded without the user directly choosing, because the browser was closed. Grouped by the file extension.

User chose to discard a download which was marked DANGEROUS_FILE. Grouped by the file extension.

Reason for dropping a record read in from the DB.

Number of downloads removed from the history at once.

How long it took to delete some downloads from history.

How long it took to delete some downloads from history, per download.

A download which was marked dangerous was discarded without the user directly choosing, because the browser was closed. Grouped by the type of danger.

The percentage of the available disk bandwidth that was used by the download. 100% indicates that the disk bandwidth was the limiting factor for the download.

User actions in chrome://downloads

The size of successfully completed downloads.

A download warning was shown in the shelf. Note that some downloads may not be shown on the shelf, e.g., if chrome://downloads is already open when the download completes, or if an extension is using the downloads API. Grouped by the type of danger.

How the user interacts with the file chooser when doing a "Save As" for non-full-page saves.

The amount of time in milliseconds the file thread blocks for each set of buffers drained from the incoming pipe (ms).

The number of buffers in a call to DownloadManager::UpdateDownload.

The time between a download completing and the file being opened for the first time.

The number of items in the History database, at the time a new download is recorded.

The number of items in the History database, at the time a new download is recorded. Higher maximum, more buckets than Download.HistorySize.

Positive net error code that caused a download to be interrupted at the *end* of a download (when the number of bytes is known). This is only triggered when the total content size is known before any bytes are transferred, such as when a Content-Length header is supplied.

The reason that a download was interrupted at the *end* of a download (when the number of bytes is known). This is only triggered when the total content size is known before any bytes are transferred, such as when a Content-Length header is supplied.

Positive net error code that caused a download to be interrupted.

The excessive number of bytes which have been received at the time that a download is interrupted. This is only triggered when the total content size is known before any bytes are transferred, such as when a Content-Length header is supplied.

The reason that a download was interrupted.

The number of kilobytes received for a download at the time it is interrupted.

The reported total size in kilobytes for a download at the time it is interrupted. This is essentially the size reported by the Content-Length header. If no size is specified up-front, it is not recorded in the histogram. For example, a download transferred with chunked encoding will not be recorded.

The total number of bytes minus the received number of bytes at the time that a download is interrupted. This is only triggered when the total content size is known before any bytes are transferred, such as when a Content-Length header is supplied.

True if the size of an interrupted download is unknown, false if it is known.

A download has been marked as malicious. Grouped by the type of danger. Each download can only be recorded once; it will be labeled with the first type of danger spotted.

Network error that produced a DOWNLOAD_INTERRUPT_REASON_NETWORK_FAILED result in DownloadResourceHandler::OnResponseCompleted().

Windows error that produced a DOWNLOAD_INTERRUPT_REASON_ACCESS_DENIED result in MapShFileOperationCodes().

Windows error that produced a DOWNLOAD_INTERRUPT_REASON_FILE_FAILED result in MapShFileOperationCodes().

Percentage of DownloadItem::Observer::OnDownloadUpdated events that signified a change in the extension API representation of the download.

Invocation count for methods of opening a download. For some file types, Chrome defaults to opening the file in the browser instead of invoking the system handler. The user has the option of overriding this behavior.

The number of unopened downloads, when one is opened.

The time between a download completing and the file being opened.

Changes observed when a response is received for a full download resumption request.

Changes observed when a response is received for a partial (byte-range) download resumption request.

The maximum bandwidth (per read) that Chrome could have provided for the download. If the actual bandwidth equals the potential bandwidth, that means that Chrome was the limiting factor for download bandwidth.

The percentage of the lifetime of the DownloadResourceHandler for which it was blocked by downstream flow control. 0% indicates that the network bandwidth was the limiting factor for the download.

Events (e.g. Started, Cancelled, Finished, Write to Completed file, Write to Failed file) occuring within the state machine of a SavePackage operation.

The number of download items in progress on the shelf when it closes automatically.

The number of download items in progress on the shelf when the user closes it.

The number of download items on the shelf when it closes automatically.

The number of download items on the shelf when the user closes it.

User saw the confirm prompt to save a download which was marked dangerous. Grouped by the type of danger.

The initiation source (if initiated within the content layer of chrome) for a download.

The initiation source (if initiated within the above-content layer of chrome) for a download.

Time between the start of a download and its completion.

User chose to discard a download which was marked dangerous. Grouped by the type of danger.

The number of iterations for the write loop in BaseFile::AppendDataTofile().

The write size for calls to BaseFile::AppendDataTofile().

Deperecated 8/2013.

Status of drive cache metadata database open.

Time spent to load the list of files in a single directory from Google Drive server.

Deprecated 12/2013 due to the UMA stat bucket layout change. We'll use Drive.FullFeedLoadTime instead.

Time spent to load the entire file system information from the server

Deprecated 10/2012.

Provides breakdown of specific formats for hosted documents. Recorded when feed is loaded from the server.

Deprecated 10/2012.

Provides breakdown of specific file formats for regular files. Recorded when feed is loaded from the server.

Time spent to load the entire file system information from the server

Deperecated 12/2013 since it did not record meaningful information. Drive.DirectoryFeedLoadTime should be checked for measuring the time until the user sees the first response of file lists.

Time spent to load the initial part of the file system information from the server

Result of drive resource metadata database initialization.

Result of attempt to open existing drive resource metadata database.

Version number of drive resource metadata DB found on the disk before checking whether it should be upgraded. Recorded during Drive metadata initialization triggered by profile initialization.

Number of files recovered from Drive cache directory. Recorded when file recovery takes place after metadata DB corruption is found during metadata DB initialization.

Number of hosted documents (spreadsheets etc.) on Drive. Logged when Drive is first accessed.

Number of regualr files on Drive. Logged when Drive is first accessed.

Number of total files (regualr files + hosted documents) on Drive. Logged when Drive is first accessed.

Tracks whether the push notification is initially enabled for Drive. Recorded when the first notification is processed. Notification is emulated by polling if the push notication is disabled.

Tracks whether the push notification request is registered correctly for Drive. Recorded when the push notification manager is initialized.

Time spent to perform an incremental search for auto completion of files on Drive. This time is collected for every partial query the user types for auto completion. For instance, if the user types "faq", incremental searches are performed for "f", "fa", and "faq" respectively.

Time since the user logged in until the auto-enrollment protocol completed. 0 is sampled when the protocol is done by the time the user logs in.

Total duration time of the auto-enrollment protocol.

Network error code (if applicable) for auto-enrollment requests.

URL fetcher status for auto-enrollment requests.

Events related to fetching, saving and loading DM server tokens. These are used to retrieve cloud policies.

Events related to device enrollment on new installs of ChromeOS devices.

Result of the OpenNetworkConfiguration policy validation.

A set of enterprise policy rules that are in use. This is recorded every 24 hours and at startup, if the last recording was earlier than a day before.

Events related to fetching, saving and loading user policies, and also device policies on ChromeOS.

Events for counting policy invalidations received with and without payloads. Invalidations indicate that a policy has been updated and should be refreshed. Payloads provide context about the policy update, but may be absent if dropped by the invalidation service.

Time since startup of the cloud policy code until the policy invalidation service first reported its online status.

Load status from the policy loaders which pull policy settings from the underlying platform, such as Windows Group Policy.

Events measuring effectiveness of refreshing policy when invalidations are received from a service. For each refresh, indicates whether the policy changed, and whether the policy was invalidated at the time of the refresh.

Initialization delay due to loading the user policy cache.

Policy client error during initial policy fetch.

Delay for registering the client with the policy server.

Delay for minting an OAuth2 acccess token.

Delay for fetching policy from the policy server.

Total delay for the initial policy fetch.

Service error during OAuth2 access token fetch.

Network error during OAuth2 access token fetch.

Number of Mouse events coalesced.

Number of Touch events coalesced.

Time between the first and last events in a coalesced mouse events group.

Time between the first and last events in a coalesced touch events group.

Time between initiation of all input events and browser processing.

Time between initiation of input event and browser processing.

Time between touch events sent from RWH to renderer and acked by renderer.

Time between touch events received by Chrome and sent from RWH to renderer.

Time between initiation of all input events and renderer processing.

Time between initiation of input event and browser processing.

Deprecated 12/2013 and replaced by Event.Latency.RendererImpl.GestureScroll2

Time between initial creation of touch event and when the resulting ScrollGesture reaches Impl thread. Maximum is 200ms.

Time between initial creation of touch event and when the resulting ScrollGesture reaches Impl thread. Maximum is 1000ms. This supersedes the Event.Latency.RendererImpl.GestureScroll metric.

Time between initial creation of touch event and the resulting frame from ScrollUpdate is swapped.

For each pageload, the number of extensions that inject a content script.

For each pageload, the number of extensions that create divs to add to the page.

For each pageload, the number of extensions that create 'embed' elements to add to the page.

For each pageload, the number of extensions that create iframes to add to the page.

For each pageload, the number of extensions that create inputs to add to the page.

For each pageload, the number of extensions that create links to add to the page.

For each pageload, the number of extensions that create 'object' elements to add to the page.

For each pageload, the number of extensions that create script tags to add to the page.

For each pageload, the number of extensions that use document.write.

For each www.google.com pageload, the number of extensions that inject a content script.

For each www.google.com pageload, the number of extensions that create divs to add to the page.

For each www.google.com pageload, the number of extensions that create 'embed' elements to add to the page.

For each www.google.com pageload, the number of extensions that create iframes to add to the page.

For each www.google.com pageload, the number of extensions that create inputs to add to the page.

For each www.google.com pageload, the number of extensions that create links to add to the page.

For each www.google.com pageload, the number of extensions that create 'object' elements to add to the page.

For each www.google.com pageload, the number of extensions that create script tags to add to the page.

For each www.google.com pageload, the number of extensions that use document.write.

For each www.google.com pageload, the number of extensions that set innerHTML.

For each www.google.com pageload, the number of extensions that invoke DOM methods.

For each www.google.com pageload, the number of extensions that set the value of DOM properties via assignments.

For each www.google.com pageload, the number of extensions that read from the DOM.

For each pageload, the number of extensions that set innerHTML.

For each pageload, the number of extensions that invoke DOM methods.

For each pageload, the number of extensions that set the value of DOM properties via assignments.

For each pageload, the number of extensions that read from the DOM.

The number of extensions that were blacklisted when already installed, grouped by Extension::Location. Logged when ExtensionService blackists and unloads an installed extension.

The number of extensions that have been blocked from installing grouped by Extension::Location. Logged when ExtensionService refuses to install a blacklisted extension.

The number of extensions that have been silently installed in a blacklisted state, grouped by Extension::Location. Logged when ExtensionService installs a blacklisted extension without blocking it (ExtensionBlacklist.BlockCRX would be logged otherwise). Typically this will be when a user has a blacklisted extension synced.

The number of extensions that were unblacklisted when installed, grouped by Extension::Location. Logged when ExtensionService unblacklists and loads a blacklisted extension.

Counts whether we force-disabled an installed extension at startup because a policy provider indicated it must remain disabled.

Whether the server result received by the extensions install signer was valid or invalid.

The result of the verifier trying to get a new signature.

The result of initialization for the extension install verifier.

The outcome for each call to InstallVerifier::MustRemainDisabled.

Records when the allocation of IDs for chrome.runtime.Port overflows.

Captures the results of URL resolution when relative urls are used in the tabs/windows api.

The number of times v1 apps are launched grouped by extension_misc::AppLaunchBuckets. See also Apps.AppLaunch for v2 apps.

The number of times apps are launched grouped by extensions::LaunchContainer.

The number of apps loaded at startup time grouped by Extension::Location.

The actions taken in the NTP apps promo grouped by extension_misc::AppsPromoBuckets.

The number of apps launched grouped by extensions::LaunchType.

The time for an extension's background page to load.

The type (if any) of background page the extension has. Recorded for installed extensions on startup.

Net error results from URLFetcher.

Number of times chrome retried to download an extension with a url on a google.com domain, before eventually giving up.

Number of times chrome retried to download an extension with a url on a non google.com domain, before eventually giving up.

Number of times chrome retried to download an extension with a url on a google.com domain, before eventually succeeding.

Number of times chrome retried to download an extension with a url on a non google.com domain, before eventually succeeding.

Length of the path to the directory under which an extension is installed. This directory is in the user's profile.

Time spent until rules storage delegate gets ready.

Number of extensions referenced in the depricated external extensions source at path chrome::DIR_DEPRICATED_EXTERNAL_EXTENSIONS.

The time for a dialog-hosted extension to load.

The number of extensions that are disabled at browser startup.

The number of extensions that are disabled at browser startup due to permissions increases.

User response to the dialog shown when an extension is disabled due to an update requiring more permissions.

If opening the CRX file for unpacking fails, this integer is the error code given by the OS.

The time an extension's event page has spent loaded.

The time an extension's event page has spent unloaded.

The time for an extension's event page to load.

An extension has been installed.

The number of extensions loaded at startup time grouped by Extension::Location.

Length of the Extensions dir path inside the profile directory.

Time taken for the ExtensionService to initialize, including the time it takes to load the extensions for the service's profile and parse their manifests. This happens during startup and also any time a new profile is loaded.

An extension has been uninstalled.

Records what happens to extensions that are sideloaded, grouped by the ExternalExtensionEvent enum.

The number of sideloaded apps/extensions loaded on startup grouped by enabled/disabled state.

Number of extensions referenced in the external extensions source at path chrome::DIR_EXTERNAL_EXTENSIONS.

Number of apps/extensions loaded on startup with an inconsistent "from webstore" state. This means an item that is flagged as from_webstore, but with either a non-webstore update_url or an external install location.

Number of calls to extension functions.

What happens when the extensions system tries to get a temp dir to unpack in?

The amount of time for a CSS file to be injected into a page.

Number of scripts injected at document end by extensions.

Time taken to inject all scripts at document end by extensions.

Number of scripts injected at document idle by extensions.

Time taken to inject all scripts at document idle by extensions.

Time taken to inject all scripts by extensions.

Number of css files injected by extensions.

Number of scripts injected at document start by extensions.

Time taken to inject css/scripts at document start by extensions.

Installs grouped by the location property in prefs.

Installs grouped by Extension::HistogramType.

The number of extensions and themes loaded at browser startup.

Time taken to load all extensions at browser startup.

The number of apps loaded by each user at startup time.

The number of externally managed apps loaded by each user at startup time.

The number of user-installed apps loaded by each user at startup time.

The number of browser action extensions loaded at browser startup.

The number of content-pack extensions loaded at browser startup.

The number of extensions loaded at browser startup.

The number of externally managed extensions loaded at browser startup.

The number of user-installed extensions loaded at browser startup.

The number of externally managed extensions and apps loaded at browser startup.

The number of hosted apps loaded by each user at startup time.

The number of legacy packaged apps loaded by each user at startup time.

The number of page action extensions loaded at browser startup.

The number of platform apps loaded at browser startup.

The number of themes loaded at browser startup.

The number of extensions loaded at startup time grouped by Extension::HistogramType.

The number of converted user scripts loaded at browser startup.

Number of times chrome retried to download an extension update manifest with a url on a google.com domain, before eventually giving up.

Number of times chrome retried to download an extension update manifest with a url on a non google.com domain, before eventually giving up.

Number of times chrome retried to download an extension update manifest with a url on a google.com domain, before eventually succeeding.

Number of times chrome retried to download an extension update manifest with a url on a non google.com domain, before eventually succeeding.

Number of extension loads on startup where it is necessary to reload the mainfest because the locale has changed.

Number of extension loads on startup where it is not necessary to reload the extension's manifest.

Number of extension loads on startup where it is necessary to reload the manifest because the extension is unpacked.

Time that network requests were blocked due to extensions.

Percentage of total lifetime a network request was blocked due to an extension.

Time that network requests were blocked due to relevant rule registries loading.

The number of apps/extensions with a non-webstore update_url loaded at startup time grouped by Extension::Location.

Number of non-WebStore extensions on startup that override the new tab page.

The permissions present in an extension when it is automatically disabled due to a permission increase (e.g., after an extension upgrade).

The permissions present in an extension when it was installed.

The permissions present in an extension when installation was aborted, not including installation errors and user cancels.

The permissions present in an extension when installation was canceled.

The permissions present in an extension when it was loaded.

The permissions present in an extension when it was re-enabled from a confirmation prompt.

The permissions present in an extension when the re-enable prompt was aborted, not including installation errors and manual user cancels.

The permissions present in an extension when the re-enable was canceled from the confirmation prompt.

The permissions present in an extension when it was uninstalled.

The permissions present in an extension when it was installed through the web store.

The permissions present in an extension when installation from the web store was aborted, not including installation errors and user cancels.

The permissions present in an extension when installation from the web store was canceled.

The initialization latency (in milliseconds) introduced to each extension resource request by querying the directory timestamp.

The difference in seconds between an extension resource's last modification time and its extension base directory's creation time. Recorded on each extension resource request if the difference is non-negative (i.e., the resource's last modification time is more recent than the directory's creation time.) For cases where the directory creation date is more recent, see Extensions.ResourceLastModifiedNegativeDelta instead.

The absolute difference in seconds between an extension resource's last modification time and extension base directory's creation time. Recorded on each extension resource request if the difference is negative (i.e., the directory's creation time is more recent than the resource's last modification time.) For cases where the resource modification time is more recent, see Extensions.ResourceLastModifiedDelta instead.

Count the number of times a sandboxed extension unpack fails.

What caused a sandboxed extension unpack to fail?

Time taken to unpack an extension, when the unpack fails.

Length of the initial path to the CRX to be unpacked.

Length of the normalized (link/junction free) path to the temporary copy of a CRX made during unpacking.

Rate at which a CRX file is unpacked in Kilobytes per second.

Rate at which CRX files 1MB to 2MB are unpacked in Kilobytes per second.

Rate at which CRX files 2MB to 5MB are unpacked in Kilobytes per second.

Rate at which CRX files 50kB to 1MB are unpacked in Kilobytes per second.

Rate at which CRX files 5MB to 10 MB are unpacked in Kilobytes per second.

Rate at which CRX files larger than 10MB are unpacked in Kilobytes per second.

Rate at which CRX files under 50 KB are unpacked in Kilobytes per second.

Count the number of times a sandboxed CRX unpack succeeds.

Count the number of times a sandboxed CRX unpack succeeds, but we can't get the file size.

Time taken to unpack an extension, when the unpack succeeds.

Length of the path of the temporary copy of a CRX made during unpacking.

Length of the path under which a CRX is unpacked.

The time one extension delays network requests at startup.

The total time extensions delay network requests at startup.

Time taken to load a toolstrip.

Uninstalls grouped by Extension::HistogramType.

Count failing CRX installs, grouped by the way an extension can be installed.

Count successful CRX installs, grouped by the location property in prefs. installed.

Count successful CRX installs, grouped by the cause of the install.

Count successful CRX installs, grouped by the location property in prefs.

The number of legacy packaged apps and hosted apps that were checked during an update check.

The number of extensions that were checked during an update check.

Time in minutes between update checks.

The number of crx's with a Google-hosted update URL that were checked during an update check.

The number of crx's with no update URL checked during an update check.

The number of crx's with a non-Google update URL that were checked during an update check.

The number of packaged apps that were checked during an update check.

The number of themes that were checked during an update check.

The number of extensions that were updated at browser startup.

Deprecated 10/2013.

What happened when the extension updater tried to write a file?

Updates grouped by the location property in prefs.

Updates grouped by Extension::HistogramType.

The reason a webstore download was interrupted.

The number of KBytes received for a webstore download before it was interrupted.

The total expected size in KBytes of an interrupted webstore download.

Tracks whether the total size of an interrupted webstore download was known.

Chrome OS File Browser opening mode.

Chrome OS File Browser: time to scan a directory. Measured on every File Browser directory change.

Tracks whether download destination is set to a Google Drive folder when the download destination is changed by the user in the settings page.

Tracks whether download destination is set to a Google Drive folder on startup.

Chrome OS File Browser: number of files and directories in the Downloads directory (not including the contents of nested directories). Computed every time the File Browser current directory changes to Downloads.

Chrome OS File Browser: this is recorded when the user adds a folder shortcut.

Chrome OS File Browser: number of saved folder shorcuts. This is recorded when Files.app is launched.

Chrome OS File Browser: this is recorded when the user clicks or selects a folder shortcut and is navigated to the target folder.

Chrome OS File Browser: this is recorded when the user removes a folder shortcut.

Chrome OS File Browser is an built-in extension without a background page. Its main.html file is loaded every time the user opens a File Browser tab or a file chooser dialog. The file is fairly large and the initialization is pretty expensive.

Deprecated 4/2013, and replaced by FileBrowser.ViewingFileType.

File types that were tried to be opened through browser.

Chrome OS Photo Editor: time to display an image. Measured from the moment the user selected the image till the moment it is displayed (not counting the low resolution preview).

Chrome OS Photo Editor: the type of the file opened.

Chrome OS Photo Editor: the way the image has been loaded.

Chrome OS Photo Editor: time to load an image from a file.

Chrome OS Photo Editor: the result of a file save operation.

Chrome OS Photo Editor: time to save an image to a file.

Chrome OS Photo Editor: size of an image file in megabytes. Measured on every image load.

Chrome OS Photo Editor: size of an image in megapixels. Measured on every image load.

Chrome OS Photo Editor: the button which the user clicked.

Chrome OS Photo Import flow: action chosen in the Action Choice dialog for the external device.

Chrome OS Photo Import flow: the number of photos imported. Measured on every successfull import operation.

Chrome OS Photo Import flow: the percent of photos imported among all the photos on the device. Measured on every successfull import operation.

Chrome OS Photo Import flow: time to load the action dialog. Measured between the moment window appears and the moment user see all available actions for the device.

Chrome OS Photo Import flow: time to scan the external device.

Chrome OS File Browser: the reason why the suggest apps dialog was closed.

Chrome OS File Browser: whether the Webstore item user selected was successfully installed or not.

Chrome OS File Browser: whether the initialization of the dialog succeeded or not.

Chrome OS File Browser: time to load the suggest apps dialog. Measured between the moment window appears and the moment all the contants in the dialog including the Chrome Webstore widget are ready.

File types that were tried to be viewed through browser. This is recorded when the user tries to view a file from Files.app.

Result of the authentication for Drive.

Deprecated 9/2012, and replaced by Drive.EntireFeedLoadTime

Time spent to load the entire file system information from the server

Deprecated 9/2012, and replaced by Drive.EntryKind

Provides breakdown of specific formats for hosted documents. Recorded when feed is loaded from the server.

Deprecated 9/2012, and replaced by Drive.InitialFeedLoadTime

Time spent to load the initial part of the file system information from the server

Deprecated 9/2012, and replaced by Drive.NumberOfHostedDocuments

Number of hosted documents (spreadsheets etc.) on Drive. Logged when Drive is first accessed.

Deprecated 9/2012, and replaced by Drive.NumberOfRegularFiles

Number of regualr files on Drive. Logged when Drive is first accessed.

Deprecated 9/2012, and replaced by Drive.NumberOfTotalFiles

Number of total files (regualr files + hosted documents) on Drive. Logged when Drive is first accessed.

Whether high accuracy geolocation information was requested.

Events in GeolocationInfoBarDelegate.

Events in GeolocationInfoBarDelegateAndroid.

The number of wifi access points used to determine geolocation.

Events in NetworkLocationRequest.

Http response codes in NetworkLocationRequest.

Events in Google Now component extension.

Refresh rate of the display in Hz. This is recorded every time we present a frame.

The time that the GPU process spends collecting driver information during startup.

The time that the browser process takes to create the compositor from its point of view. One of these is created for each top-level window (browser frame, menus, etc.).

The time that the GPU process spends in initializing the GL surface, and collecting graphics information.

Count of "History Index *" databases deleted. These databases stored the full-text-search data for history, which was removed at r213442, this histogram tracks cleanup.

Size of the recovered Favicons database relative to the original corrupt database. Recovery is VACUUM-like, so the resulting database should always be smaller. Substantial 100% results would indicate empty databases being recovered, substantial low% results would indicate very little data being recovered.

Rows recovered from [favicon_bitmaps] table in Favicons recovery.

Rows recovered from [favicons] table in Favicons recovery.

Rows recovered from [icon_mapping] table in Favicons recovery.

Track results of SQLite database recovery code in thumbnail_database.cc.

Size of the recovered TopSites database relative to the original corrupt database. Recovery is VACUUM-like, so the resulting database should always be smaller. Substantial 100% results would indicate empty databases being recovered, substantial low% results would indicate very little data being recovered.

Rows recovered from [thumbnails] table in TopSites recovery.

The TopSites recovery code is written conservatively, with successful recovery committed and any failure leading to rollback. This tracks the outcomes to determine which cases are high-frequency enough to warrant adding additional code to handle them (versus simply deleting the data).

Page visits to each of a user's top 50 sites. Visits to all other sites go into the 51st bucket. Only count the page visit if it came from user browsing and only count it once when cycling through a redirect chain.

Number of entries that the clicked entry is older than in History page. Last bucket is any entry of that value or higher.

Subset of the Click Position histogram. Contains only the first smaller subset of entries on the page. Number of entries that the clicked entry is older than in History page. Last bucket is entries of that value or higher.

Histogram for usage of the section in the history page that allows the user to access tabs from other devices.

Number of entries that the deleted entry is older than in History page. Last bucket is any entry of that value or higher. Confirmed removal is not guaranteed, just an initiation of 'Remove selected items'.

Subset of Remove Entry Position histogram. Contains only the first smaller subset of entries on the page. Number of entries that the deleted entry is older than in History page. Last bucket is any entry of that value or higher. Confirmed removal is not guaranteed, just an initiation of 'Remove selected items'.

The time spent waiting for write lock on a disk cache entry.

Result of a main page HttpCacheTransaction if offline mode had been enabled.

Net error results from non-restartable cache read errors.

Net error results from restartable cache read errors.

The type of Vary header for a given GET response.

The importer used on first run Auto Import.

The importer used on import from the bookmarks file API.

The importer used on import from the chrome://settings/importData UI.

The amount of time from install time to time that user opens import dialog from BookmarkBarView.

The amount of time from install time to time that user opens import dialog from NTP floating BookmarkBarView.

Deprecated and replaced by Import.ShowDialog.FromBookmarkBarView

The amount of time from install time to time that user opens import dialog from BookmarkBarView.

Deprecated and replaced by Import.ShowDialog.FromFloatingBookmarkBarView

The amount of time from install time to time that user opens import dialog from NTP floating BookmarkBarView.

Whether the Sensor.TYPE_LINEAR_ACCELERATION was available at the start of Device Motion.

Whether the Sensor.TYPE_ACCELEROMETER was available at the start of Device Motion.

Whether the SENSOR_TYPE_ACCELEROMETER_3D was available at the start of Device Motion on the Windows platform.

Whether the SENSOR_TYPE_GYROMETER_3D was available at the start of Device Motion on the Windows platform.

Whether the Sensor.TYPE_GYROSCOPE was available at the start of Device Motion.

Whether the SENSOR_TYPE_INCLINOMETER_3D was available at the start of Device Orientation on the Windows platform.

This histogram counts the number of Device Motion API invocations in the default implementation (Linux and CrOS). The default implementation does not provide any sensors so the result is always false.

Whether the sudden motion sensor was available at the start of Device Motion on the MacOS platform.

This histogram counts the number of Device Orientation API invocations in the default implementation (Linux and CrOS). The default implementation does not provide any sensors so the result is always false.

Whether the sudden motion sensor was available at the start of Device Orientation on the MacOS platform.

Whether the Sensor.TYPE_ROTATION_VECTOR was available at the start of Device Orientation.

The number of update attempts until the update has been applied. This is reported every time the device has completed an update.

Errors from update_engine process when running in dev mode.

The overhead in downloading extra bytes due to errors/interruptions. Expressed as a percentage of the bytes that are actually needed to be downloaded for the update to be successful.

The combinations of protocol and source server that were used to complete a successful update.

The number of update attempts with a full update payload until the update has been applied. This is reported on every update attempt.

Errors from update_engine process when running in normal mode.

The age of the OS, defined as the age of the /etc/lsb-release file. This is reported on every update check but at most once a day.

The number of update attempts until the update has been applied. This is reported on every update attempt.

The type of update payload used to update the device. The difference between "Full" and "Forced Full" is that in the latter, the request sent to Omaha included a directive saying that a delta payload wasn't accepted. A "Full" payload is one where a delta payload was accepted but Omaha provided a full payload. This is reported every time the device has completed an update.

The number of consecutive times a device has failed to boot an update that successfully applied. This metric is reported every time the firmware fails to boot the slot with the update and fell back to the slot it originally updated from.

Number of MBs downloaded from during an update that completed successfully.

Wall-clock duration between when an update has successfully completed (and the user is presented with the "reboot arrow") and when the system has booted into the new update. This is reported every time the device is rebooted after an update has been applied.

Total number of MBs downloaded since the last successful update. This also includes all the bytes downloaded during any prior failed attempts.

Absolute wall-clock time duration it took for the update to complete from the time an update first began. It includes not just the time the device was up, but also includes the time the device spent sleeping.

Uptime duration it took for the update to complete from the time an update first began. It does not include the time the device spent sleeping, but it does include the uptime spent in waiting for the hourly update checks to happen.

Number of times the device was rebooted by the user since an update began and until it completed successfully.

The number of update attempts that didn't complete because a newer update was detected during the update operation. This is reported every time the device has completed an update.

The number of consecutive different abandoned update payloads since the last successful update. This is reported every time an update payload is abandoned because a newer update payload is available.

Number of times the download URLs were switched due to failures.

Records various events of interest in the InstantController. E.g. When URLs are blacklisted.

How often an Instant preview is committed onto a different tab than it was created from.

The time between the first Omnibox interaction and when the Instant preview shows. If the instant preview was already showing when the user interacted with the omnibox, this histogram is not recorded.

Records a histogram for how often the Cacheable NTP fails to load.

Records the cause for falling back to a local overlay at the time of fallback.

Deprecated as of 10/2013. This histogram is no longer relevant since the HTML overlay went away.

Records a histogram for instant extended (Local NTP and Online NTP) and non-extended navigations.

Deprecated as of 11/2013.

Records, on startup, whether the user has chosen to opt-in to or opt-out of InstantExtended via chrome://flags.

Deprecated 2013-06. As of m30 use InstantExtended.NewOptInState.

Records, on startup, whether the user has chosen to opt-in to or opt-out of InstantExtended via chrome://flags.

Deprecated 2013-07. Please see InstantExtended.PercentageMatchV2_QuerytoQuery instead.

Records the number of matching characters at the start of the user's text as a percentage of average length between the old and new text when the user navigates from a search query to another search query.

Deprecated 2013-07. Please see InstantExtended.PercentageMatchV2_QuerytoURL instead.

Deprecated 2013-07. Please see InstantExtended.PercentageMatchV2_URLtoQuery instead.

Records the number of matching characters at the start of the user's text as a percentage of average length between the old and new text when the user navigates from a url to a search query.

Deprecated 2013-07. Please see InstantExtended.PercentageMatchV2_URLtoURL instead.

Records the number of matching characters at the start of the user's text as a percentage of average length between the old and new text when the user navigates from a url to another url.

Records the number of matching characters at the start of the user's text as a percentage of average length between the old and new text when the user navigates from a url to a search query.

Records the number of matching characters at the start of the user's text as a percentage of average length between the old and new text when the user navigates from a url to another url.

Deprecated 2013-06. This preference has not been exposed or used for months, and we do not plan to use it in the future.

Records, on startup, the value of the "Allow your search engine to provide Instant result" preference setting for the first profile loaded.

Removed on 8/1/13.

The time between the SSL interstitial display and the user decision, which may be either accept or deny. This is only recorded for overridable SSL warnings with a CERT_AUTHORITY_INVALID warning. Timing begins when user first focuses on the page.

Removed on 8/1/13.

The time between the SSL interstitial display and the user decision, which may be either accept or deny. This is only recorded for overridable SSL warnings with a CERT_COMMON_NAME_INVALID warning. Timing begins when user first focuses on the page.

Removed on 8/1/13.

The time between the SSL interstitial display and the user decision, which may be either accept or deny. This is only recorded for overridable SSL warnings with a CERT_DATE_INVALID warning. Timing begins when user first focuses on the page.

User action when the user is shown a SSL interstitial. SHOW_ALL and MORE refer to the total number of SSL errors; all of the other numbers pertain to the number of actions related to SSL errors that are overridable. The counts do not sum to 100%; SHOW_ALL is a superset of SHOW_OVERRIDABLE, which in turn will be a supserset of the PROCEED/DONT_PROCEED variables. SHOW_UNDERSTAND is only being used by an experimental field trial.

Removed on 8/1/13.

The time between the SSL interstitial display and the user decision, when the user accepts the SSL warning. This is only recorded for overridable SSL warnings. Timing begins when user first focuses on the page.

The type of SSL error that the user encounters. This is recorded for all SSL warnings, regardless of whether they are overridable.

Removed on 8/1/13.

The time between the SSL interstitial display and the user decision, when the user rejects the SSL warning. This is only recorded for overridable SSL warnings. Timing begins when user first focuses on the page.

The time between keystrokes in Aura text fields. The only keystrokes that are measured are ones that produce a printable character and are not over 5 seconds apart.

Accept languages.

Application languages used for UI.

Deprecated 2013-10. No thread-unsafety was found.

Linux and CrOS use unlocked_stdio(3). If it is used unsafely, record it here. If there is no record of unsafety after chrome 29 has been in the stable channel for a few weeks then revert this change.

Methods where leveldb's Chromium environment has IO errors when being used by IndexedDB.

PlatformFileErrors encountered by a single leveldb env method.

Errno of errors encountered in NewLogger.

Errno of errors encountered in NewSequentialFile.

Deprecated 2013-04. As of m28 use LevelDBEnv.IDB.IOError.NewRandomAccessFile.

File errors in leveldb IDBEnv's NewRandomAccessFile method.

Errno of errors encountered in WritableFileAppend.

Errno of errors encountered in WritableFileFlush.

Number of directories missing when IDB LevelDBEnv tries to create a Lock file.

File descriptor limit recorded every time LevelDB calls NewRandomAccessFile for IndexedDB.

Number of backup files found without corresponding ldb files. As measured by GetChildren when used in IndexedDB.

When IDB LevelDBEnv successfully retries an operation that had failed, record the error from the most recent failed attempt.

Success indicates a successful backup or restore operation for .ldb table files when used in IndexedDB.

Deprecated 2013-04. As of m28 use LevelDBEnv.IDB.TimeUntilSuccessFor.

Time IDB LevelDBEnv slept before successfully completing this operation. 0 means success on the first try.

Methods where leveldb's Chromium environment has IO errors.

PlatformFileErrors encountered by a single leveldb method.

Errno of errors encountered in NewLogger.

Errno of errors encountered in NewSequentialFile.

Deprecated 2013-04. As of m28 use LevelDBEnv.IOError.NewRandomAccessFile.

File errors in leveldb ChromiumEnv's NewRandomAccessFile method.

Errno of errors encountered in WritableFileAppend.

Errno of errors encountered in WritableFileFlush.

Number of directories missing when Non-IDB LevelDBEnv tries to create a Lock file.

File descriptor limit recorded every time LevelDB calls NewRandomAccessFile for clients other than IndexedDB.

Number of backup files found without corresponding ldb files. As measured by GetChildren when used in LevelDB clients other than IndexedDB.

When Non-IDB LevelDBEnv successfully retries an operation that had failed, record the error from the most recent failed attempt.

Success indicates a successful backup or restore operation for .ldb table files when used by LevelDB clients other than IndexedDB.

Deprecated 2013-04. As of m28 use LevelDBEnv.TimeUntilSuccessFor.

Time Non-IDB LevelDBEnv slept before successfully completing this operation. 0 means success on the first try.

The version of glibc used. (Linux only)

Chrome OS login failure reason.

Chrome OS histogram that keeps track of the days since the least frequently used account signed in. Reported on every boot and once a day after that.

The state of Chrome OS owner key and device policy files.

Time from first display of the login prompt until the user completes signing in.

Chrome OS login success reason.

Chrome OS histogram that keeps track of the way a user logs in and whether Chrome OS is running normal or developer mode.

Whether accelerated compositing was used for HTML5 media rendering.

Bits per channel of HTML5 audio sample data.

Audio channel layout in HTML5 media.

Audio codec used in HTML5 media.

Measures the time taken for AudioInputController::

Measures the time taken for AudioInputDeviceManager::

Measures the time taken for AudioOutputController::

Time spent waiting in AudioOutputController::WaitTillDataReady() if the data was not initially available.

Whether playback started successfully after stream startup was requested.

Captures statistics for various AudioRendererImpl events.

Percentage of AudioSyncReader::Read() calls where the renderer missed its realtime deadline.

Audio sample format in HTML5 media. Logged when Audio Decoder initializes.

Audio samples per second in HTML5 media.

Audio samples per second in HTML5 media (atypical values, in Hz).

Whether a media response might be used to satisfy a future request.

The average number of delayed and dropped frames for the ChromeCast application. Reported every 5 seconds.

The average number of displayed frames for the ChromeCast application. Reported every 5 seconds.

Time needed to pre-buffer A/V data before the actual playback for the ChromeCast application.

Time needed to buffer A/V data after an abort for the ChromeCast application.

Time needed to buffer A/V data after an underrun for the ChromeCast application.

Audio codec used in HTML5 media.

Container used for HTML5 media.

Video codec used in HTML5 media.

Measures the actions taken in the media infobar, which prompts the users for device permission.

Duration in milliseconds of HTML5 media (when known).

addKey result using the Clear Key key system.

cancelKeyRequest result using the Clear Key key system.

Decryption error event count using the Clear Key key system.

generateKeyRequest result using the Clear Key key system.

KeyAdded event count using the Clear Key key system.

KeyError event count using the Clear Key key system.

EME NeedKey event count.

addKey result using an unknown key system.

cancelKeyRequest result using an unknown key system.

Decryption error event count using an unknown key system.

generateKeyRequest result using an unknown key system.

KeyAdded event count using an unknown key system.

KeyError event count using an unknown key system.

addKey result using the Widevine key system.

cancelKeyRequest result using the Widevine key system.

Decryption error event count using the Widevine key system.

generateKeyRequest result using the Widevine key system.

KeyAdded event count using the Widevine key system.

KeyError event count using the Widevine key system.

Bits per channel of the hardware audio device which failed to open in low latency mode and required high latency fallback.

Channel count of the hardware audio device which failed to open in low latency mode and required high latency fallback.

Channel layout of the hardware audio device which failed to open in low latency mode and required high latency fallback.

Samples per second of the hardware audio device which failed to open in low latency mode and required high latency fallback.

Samples per second of the hardware audio device (atypical values, in Hz) which failed to open in low latency mode and required high latency fallback.

Whether Chrome had to fallback to the high latency audio path or not.

The average number of delayed and dropped frames for the Fling application. Reported every 5 seconds.

The average number of displayed frames for the Fling application. Reported every 5 seconds.

Time needed to pre-buffer A/V data before the actual playback for the Fling application.

Time needed to buffer A/V data after an abort for the Fling application.

Time needed to buffer A/V data after an underrun for the Fling application.

Results of attempts to GpuVideoDecoder::Initialize().

Bits per channel of the hardware audio device.

Channel count of the hardware audio device.

Channel layout of the hardware audio device.

Samples per second of the hardware audio device.

Samples per second of the hardware audio device (atypical values, in Hz).

The time it takes to perform redirect tracking and a CORS access check while preparing to play a media file.

Audio IO layer used by the Linux OS, sampled once at startup of the browser.

State of the WebRtc local renderer, sampled once during the lifetime of a local renderer.

Audio codec used in Media Source Extensions playback. Set when AddId() is called during playback.

Number of tracks specified to AddId() for Media Source Extensions playback. May be called multiple times per element if playback is dynamically altered.

Whether Media Source Extensions is specified for playback of Media elements. Sampled when media pipeline starts.

Video codec used in Media Source Extensions playback. Set when AddId() is called during playback.

The audio bit rate as reported by the Netflix application. May be reported multiple times as network conditions change during playback.

The number of audio channels as reported by the Netflix application. May be reported multiple times as network conditions change during playback.

The average number of delayed and dropped frames for the Netflix application. Reported every 5 seconds.

The average number of displayed frames for the Netflix application. Reported every 5 seconds.

Video bit rate as reported by the Netflix application. May be reported multiple times as network conditions change during playback.

Video height as reported by the Netflix application. May be reported multiple times as network conditions change during playback.

Counts of video decode errors reported to plugin.

Number of PictureBuffers/textures requested per hardware decoder creation. This value varies by platform and video. A user visible video may trigger multiple decoder creations (sometimes every 5 seconds) but would normally not hold more than 2 sets of buffers at any given time in memory.

Vertical video resolution rounded to the nearest bucket. (Corresponds roughly to the number in 720p.)

The average number of delayed and dropped frames for the PlayMovies application. Reported every 5 seconds.

The average number of displayed frames for the PlayMovies application. Reported every 5 seconds.

Counts of video decode errors reported to RTCVideoDecoder.

Renamed to Media.RTCVideoDecoderInitDecodeSuccess.

Results of attempts to RTCVideoDecoder::InitDecode().

Indicates whether we were successful in initializing hardware video decoder for use in the RTC pipeline.

Indicates whether we were successful in initializing hardware video encoder for use in the RTC pipeline.

Video codec profile used in RTC video encoder.

Time in milliseconds from HTML5 media pipeline creation to playing event.

Size of HTML5 media (when known), in MB.

Reasons a media response won't be used to satisfy a future request.

URL scheme used with HTML5 media. (each URL provides one sample)

Error codes reported by video decode using VA-API hardware video decoder.

Measures the time taken for VideoCaptureManager::

Video codec used in HTML5 media.

Video codec profile used in HTML5 media.

Coded aspect ratio of HTML5 video.

Coded width of HTML5 video.

Pixel format color range of HTML5 video. Emitted on video load.

Pixel format used in HTML5 video. Emitted on video load.

Visible aspect ratio of HTML5 video.

Visible width of HTML5 video.

The average number of delayed and dropped frames for the YouTube application. Reported every 5 seconds.

The average number of displayed frames for the YouTube application. Reported every 5 seconds.

Time needed to pre-buffer A/V data before the actual playback for the YouTube application.

Time needed to buffer A/V data after an abort for the YouTube application.

Time needed to buffer A/V data after an underrun for the YouTube application.

Various usage counts for media galleries.

TBD.

The private working set used by the browser process. Recorded once per UMA ping.

TBD.

The private working set used by each chrome:// renderer process. Each process provides one sample. Recorded once per UMA ping.

The count of active chrome:// processes. Recorded once per UMA ping.

The private working set used by each extension process. Each process provides one sample. Recorded once per UMA ping.

The number of glyph pages present in the renderer when it commits a load. Since this is per-sub-process, you can get the average number of glyph pages in the system by multiplying this number with the average number of renderers. Note that this typically won't count the glyph pages added as a result of the load that just committed, since layout will happen after the commit. There are 512 bytes per glyph page, but this number also very closely approximates the number of glyph width map pages in the same renderer. The only difference is that if you have font fallback, it will make a new glyph page and no width page, but in most common cases there is no fallback). Width pages are 1K each (256 floats), so you could think of this value as being the number of "1.5K units related to glyphs per renderer per page load".

The private working set used by the GPU process. Recorded once per UMA ping.

System-wide graphics driver memory consumption. Recorded on Chrome OS for platforms where it is exposed by the kernel (for example, Intel i915 and Exynos Mali). Recorded once per UMA ping.

The private working set used by each Native Client loader process. Each process provides one sample. Recorded once per UMA ping.

The private working set used by each Native Client broker process. Each process provides one sample. Recorded once per UMA ping.

The count of other various utility processes (nacl, gpu, sandbox, zygote, utility). Recorded once per UMA ping.

The private working set used by each Pepper plugin process. Each plugin process provides one sample. Recorded once per UMA ping.

The private working set used by each Pepper plugin broker process. Each process provides one sample. Recorded once per UMA ping.

The count of Pepper plugin broker processes, recorded once per metrics services (UMA) update. See MetricsReportingScheduler for details.

The count of active Pepper plugin processes. Recorded once per UMA ping.

The private working set used by each plugin process. Each plugin process provides one sample. Recorded once per UMA ping.

The count of active plugin processes. Recorded once per UMA ping.

The count of all active processes. Recorded once per UMA ping.

The current process limit. Recorded once per UMA ping.

The private working set used by each renderer process. Each renderer process provides one sample. Recorded once per UMA ping.

The count of active renderer processes. Recorded once per UMA ping.

The private working set used by each sandbox helper process. Each sandbox helper process provides one sample. Recorded once per UMA ping.

The swap used by the browser process. Recorded once per UMA ping if the system has swapped.

The swap used by each chrome:// renderer process. Each process provides one sample. Recorded once per UMA ping if the system has swapped.

The amount of memory that swap was compressed into. Recorded once per UMA ping if the system has swapped.

The ratio of swapped data original size to compressed size. Recorded once per UMA ping if the system has swapped.

The swap used by each extension process. Each process provides one sample. Recorded once per UMA ping if the system has swapped.

The swap used by the GPU process. Recorded once per UMA ping if the system has swapped.

Indicates that the system has swapped memory out at least once since boot. Recorded once per UMA ping.

The amount of memory that is used by swap, including bookkeeping. Recorded once per UMA ping if the system has swapped.

The swap used by each Native Client loader process. Each process provides one sample. Recorded once per UMA ping if the system has swapped.

The swap used by each Native Client broker process. Each process provides one sample. Recorded once per UMA ping if the system has swapped.

The number of reads from swap. Recorded once per UMA ping if the system has swapped.

The number of writes to swap. Recorded once per UMA ping if the system has swapped.

The amount of memory that was swapped out. Recorded once per UMA ping if the system has swapped.

The swap used by each Pepper plugin process. Each plugin process provides one sample. Recorded once per UMA ping if the system has swapped.

The swap used by each Pepper plugin broker process. Each process provides one sample. Recorded once per UMA ping if the system has swapped.

The swap used by each plugin process. Each plugin process provides one sample. Recorded once per UMA ping if the system has swapped.

The swap used by each renderer process. Each renderer process provides one sample. Recorded once per UMA ping if the system has swapped.

The swap used by each sandbox helper process. Each sandbox helper process provides one sample. Recorded once per UMA ping if the system has swapped.

The sum of all processes' swap. Recorded once per UMA ping if the system has swapped.

The swap used by each utility process. Each utility process provides one sample. Recorded once per UMA ping if the system has swapped.

The swap used by each worker process. Each worker process provides one sample. Recorded once per UMA ping if the system has swapped.

The sum of all processes. Recorded once per UMA ping.

The private working set used by each utility process. Each utility process provides one sample. Recorded once per UMA ping.

The private working set used by each worker process. Each worker process provides one sample. Recorded once per UMA ping.

TBD.

Value of getMemoryClass() recorded once upon startup. This is an integer, device-specific constant correlated with the amount of memory available on Android device.

Reasons behind evictions of individual tabs, recorded upon each tab eviction.

Number of loaded (memory-resident) tabs when LowMemory notification is delivered.

Time between two consecutive LowMemory notification in one foreground session.

Memory notifications delivered through system callbacks to Chrome while in the background.

Memory notifications delivered through system callbacks to Chrome while in the foreground - we count LowMemory notification vs particular levels of TrimMemory foreground notification.

The result (e.g. success or the type of failure) of a modem interface switch operation performed by mist on Chrome OS.

Tracks mouse sensitivity setting changes by the user. This replaces the old Mouse.Sensitivity.Changed metric.

Tracks mouse sensitivity setting on startup. This replaces the old Mouse.Sensitivity.Started metric.

Deprecated as of 6/2013, replaced by Mouse.PointerSensitivity.Changed.

Tracks mouse sensitivity setting.

Deprecated as of 6/2013, replaced by Mouse.PointerSensitivity.Started.

Tracks mouse sensitivity setting on startup.

Measures the time elapsed between when the user mousedown-ed a link and when the user clicked a link.

For each click handled by an HTML anchor tag link, whether Blink saw a mousedown event preceding it. This is only measured for clicks handled by the anchor tag's default click event handler.

The number of mousedown events detected at HTML anchor-tag links' default event handler.

Measures the time elapsed between when the user mouseover-ed a link and when the user clicked a link.

Measures the time elapsed between when the user mouseover-ed a link and when the user mouseout-ed a link without click.

The number of mouseover events detected at HTML anchor-tag links' default event handler.

The tap gesture events detected before click at HTML anchor-tag links' default event handler.

Measures the time elapsed between when the user tapdown-ed a link and when the user clicked a link.

The number of gesturetapdown events detected at HTML anchor-tag links' default event handler.

The number of gesturetapunconfirmed events detected at HTML anchor-tag links' default event handler.

The relation of discarded tabs vs. the amount of simultaneous users. The counts are the number of discards and the buckets are the number of users. Since the count values are absolute numbers, they need to be normalized before use - so divide the counts by the percentage of users found under 'MultiProfile.UsersPerSession'.

The session counter for different multi profile modes which gets stored once per session at the beginning of the session.

Count the number of times each UI path is taken for signing into a new account in a ChromeOS multiprofile session. UI paths include the system tray and the user account switcher on the browser frame.

Count the number of times each UI path is taken for switching the active account in a ChromeOS multiprofile session. UI paths include the system tray and the keyboard shortcut.

Counts the number of window teleportations when using separated desktop mode.

Counts the number of teleported windows by types in separated desktop mode.

The number of users simultaneously signed into a multiprofile session on ChromeOS. This is recorded whenever a new user logs in.

The scheme of the URL for each main-frame navigation.

The count of unacknowledged ResourceMsg_DataReceived messages. This message is sent once per chunk of data read from the network.

The count of unacknowledged ResourceMsg_DataReceived messages at the point where we pause network loading.

The size of a SharedIOBuffer allocation.

The number of bytes copied into a SharedIOBuffer.

The percentage of a SharedIOBuffer allocation that is actually used.

The time to generate a Basic HTTP authentication token.

The time to generate a Digest HTTP authentication token.

The time to generate a Negotiate (or SPNEGO) HTTP authentication token.

The time to generate an NTLM HTTP authentication token.

Whether the certificate common name was used for matching the hostname, instead of the subjectAlternativeName. Measures results for all CAs (internal and publicly-trusted).

Whether the certificate common name was used for matching the hostname, instead of the subjectAlternativeName. Measures results ony for internal (non-publicly-trusted) CAs.

Renamed to Net.PublicKeyPinSuccess 28 Oct 2011.

A validated certificate chain may be subject to additional "pinning" requirements on a per-domain basis. This records the fraction of successful matches between a certificate chain and a pin list.

The actual amount of time spent verifying a certificate using the underlying cryptographic APIs. Because parallel verifications for the same certificate may be coalesced, histograms such as Net.SSLCertVerificationTime may be skewed, due to later verifications taking less overall time. This records the overall time spent verifying a single request, regardless of how many parallel requests are being served by the verification.

The number of times we sent N packets, but could have sent N-1 packets.

The amount of time taken before we failed to resolve the Comodo test DNS record. This is an experiment, run in conjuction with Comodo, to test the viability of a DNS based certificate revocation mechanism.

The amount of time taken to successfully resolve the Comodo test DNS record. This is an experiment, run in conjuction with Comodo, to test the viability of a DNS based certificate revocation mechanism.

The uncompressed number of bytes received per request that was compressed. Only includes requests which did not go through an explicit proxy and did not go over SSL.

The compressed number of bytes received per request that was compressed. Only includes requests which did not go through an explicit proxy and did not go over SSL.

The uncompressed number of bytes received per request that was not compressed but appears to have been compressible. Only includes requests which did not go through an explicit proxy and did not go over SSL.

The uncompressed number of bytes received per request that was compressed. Only includes requests sent through a proxy without SSL.

The compressed number of bytes received per request that was compressed. Only includes requests sent through a proxy without SSL.

The uncompressed number of bytes received per request that was not compressed but appears to have been compressible. Only includes requests sent through a proxy without SSL.

The uncompressed number of bytes received per request that was compressed. Only includes requests sent over SSL.

The compressed number of bytes received per request that was compressed. Only includes requests sent over SSL.

The uncompressed number of bytes received per request that was not compressed but appears to have been compressible. Only includes requests sent over SSL.

The count was inaccurate (it counted transactions rather than connections)

Each bucket is the number of connections of a particular type that the user has had during the session.

Renamed to match HadConnectionType.

Each bucket is the number of successful connections of a particular type that the user has had during the session.

No longer collected.

Each bucket is the number of failed connections of a particular type that the user has had during the session.

Replaced by Net.ConnectionUsedSSLVersionFallback in Chrome 21.

True if the HTTP request was to a server which requires SSLv3 fallback

Nonzero if the HTTP request was to a server which requires SSL version fallback. The value indicates the SSL version the request fell back on.

Initial typo; only here to get results from builds before r59117. See "Cookie." group.

Whether or not updates to the backing store succeeded or failed, recorded every update.

Initial typo; only here to get results from builds before r59117. See "Cookie." group.

Intervals between access time updates for each cookie.

Initial typo; only here to get results from builds before r59117. See "Cookie." group.

Number of cookies in the store (recorded every 10 minutes of active browsing time)

Initial typo; only here to get results from builds before r59117. See "Cookie." group.

For each cookie removed from the store, the reason it was removed.

Initial typo; only here to get results from builds before r59117. See "Cookie." group.

For each domain, number of cookies in that domain (recorded every 10 minutes of active browsing time).

For every top level domain, number of subdomains in that top level domain (recorded every 10 minutes of active browsing time).

Initial typo; only here to get results from builds before r59117. See "Cookie." group.

For every top level domain, number of cookies in that domain (recorded every 10 minutes of active browsing time).

Initial typo; only here to get results from builds before r59117. See "Cookie." group.

For each evicted (not expired) cookie, the amount of time since it was last used

Initial typo; only here to get results from builds before r59117. See "Cookie." group.

Number of minutes until cookie expires when set.

Initial typo; only here to get results from builds before r59117. See "Cookie." group.

The amount of time (ms) to get cookies for each URL request.

Initial typo; only here to get results from builds before r59117. See "Cookie." group.

The amount of time (ms) to load the persistent cookie store at browser start.

The total number of severs to which alternative protocol was used. This counts the number of servers persisted to prefs file.

The total number of severs that support HTTP pipelining. This counts the number of servers persisted to prefs file.

The total number of SPDY server names persisted to prefs file.

The total number of SPDY Settings properties persisted to prefs file.

When validating an HTTPS certificate we may have to block to fetch one or more revocation lists. This measures the amount of time that failures to get CRL information take.

When validating an HTTPS certificate we may have to block to fetch one or more revocation lists. This records the fraction of successful requests.

When validating an HTTPS certificate we may have to block to fetch one or more revocation lists. This measures the amount of time that each fetch takes.

The total content size in KB of all HTTP/HTTPS response bodies in the previous calendar day. The metric is reported when the first response in the current day is received.

The total content size in KB of all HTTP/HTTPS response bodies in the previous calendar day while the data reduction proxy setting was enabled. The metric is reported when the first response in the current day is received.

The total content size in KB of all HTTPS response bodies in the previous calendar day while the data reduction proxy setting was enabled. The metric is reported when the first response in the current day is received.

The total content size in KB of all long-bypassed HTTP response bodies in the previous calendar day while the data reduction proxy setting was enabled. The metric is reported when the first response in the current day is received.

The total content size in KB of all short-bypassed HTTP response bodies in the previous calendar day while the data reduction proxy setting was enabled. The metric is reported when the first response in the current day is received.

The total content size in KB of all HTTP response bodies for requests that were not served by the enabled data reduction proxy for unknown reasons in the previous calendar day while the data reduction proxy setting was enabled. The metric is reported when the first response in the current day is received.

The total content size in KB of all HTTP/HTTPS response bodies in the previous calendar day via the data reduction proxy. The metric is reported when the first response in the current day is received.

The percentage of total HTTP/HTTPS response body size while the data reduction proxy is enabled to total HTTP/HTTPS response body size in the previous calendar day. The metric is reported when the first response in the current day is received.

The percentage of total HTTPS response body size while the data reduction proxy is enabled to total HTTP/HTTPS response body size in the previous calendar day. The metric is reported when the first response in the current day is received.

The percentage of total long-bypassed response body size while the data reduction proxy is enabled to total HTTP/HTTPS response body size in the previous calendar day. The metric is reported when the first response in the current day is received.

The percentage of total short-bypassed response body size while the data reduction proxy is enabled to total HTTP/HTTPS response body size in the previous calendar day. The metric is reported when the first response in the current day is received.

The percentage of total body size of responses that were not served by the data reduction proxy for unknown reason while the data reduction proxy is enabled to total HTTP/HTTPS response body size in the previous calendar day. The metric is reported when the first response in the current day is received.

The percentage of total HTTP/HTTPS response body size via the data reduction proxy to total HTTP/HTTPS response body size in the previous calendar day. The metric is reported when the first response in the current day is received.

The percentage of data saving in the previous calendar day. A negative saving will be shown as zero. The metric is reported when the first response in the current day is received.

The percentage of data saving in the previous calendar day while the data reduction proxy was enabled. A negative saving will be shown as zero. This only counts responses while the data reduction proxy is enabled. The metric is reported when the first response in the current day is received.

The percentage of data saving in the previous calendar day via the data reduction proxy. A negative saving will be shown as zero. This only counts responses via the data reduction proxy. The metric is reported when the first response in the current day is received.

Deprecated- see Net.DailyContentLength_ViaDataReductionProxy.

Total size in KB of all response bodies in the previous calendar day that were received through the data reduction proxy.

Deprecated- see Net.DailyContentLength_DataReductionProxyEnabled

Total size in KB of all response bodies in the previous calendar day that were received when the data reduction proxy was enabled.

Deprecated- see Net.DailyContentSavingPercent.

The percentage of data saving in the previous calendar day. A negative saving will be shown as zero.

Deprecated- see Net.DailyContentSavingPercent_DataReductionProxyEnabled.

The percentage of data saving in the previous calendar day when the data reduction proxy was enabled for at least some responses during the day. A negative saving will be shown as zero.

Deprecated- see Net.DailyOriginalContentLength.

Total size in KB specified in the X-Original-Content-Length headers of all responses in the previous calendar day. If the header is not present in a response, the size of the response body is used.

Deprecated- see Net.DailyContentLength.

Total size in KB of all response bodies in the previous calendar day.

The total size in KB specified in the X-Original-Content-Length headers of all HTTP/HTTPS response bodies in the previous calendar day. If the header is not present in a response, the size of the response body is used. The metric is reported when the first response in the current day is received.

The total size in KB specified in the X-Original-Content-Length headers of all HTTP/HTTPS response bodies in the previous calendar day while the data reduction proxy is enabled. If the header is not present in a response, the size of the response body is used. The metric is reported when the first response in the current day is received.

The total size in KB specified in the X-Original-Content-Length headers of all HTTP/HTTPS response bodies in the previous calendar day via the data reduction proxy. If the header is not present in a response, the size of the response body is used. The metric is reported when the first response in the current day is received.

Deprecated- see Net.DailyContentPercent_ViaDataReductionProxy.

The percentage of Net.DailyHttpContentLengthViaDataReductionProxy in Net.DailyHttpReceivedContentLength.

Deprecated- see Net.DailyContentPercent_DataReductionProxyEnabled.

The percentage of Net.DailyHttpContentLengthWithDataReductionProxyEnabled in Net.DailyHttpReceivedContentLength.

Measures time from initiating a fetch of a PAC file from DHCP WPAD to cancellation of the fetch. For a given fetch, only one of the cancellation or completion histograms will be added to.

Measures time from initiating a fetch of a PAC file from DHCP WPAD to completion of the fetch. For a given fetch, only one of the cancellation or completion histograms will be added to.

Tracks the net error codes received when the DHCP WPAD fetch fails to retrieve a PAC file (including PAC_NOT_IN_DHCP, which is not really an error but an indication that a PAC URL was not configured in DHCP).

Tracks the frequency of each of the different known error codes of calling the GetAdaptersAddresses Win32 API.

Measures the time taken to call the GetAdaptersAddresses Win32 API, to validate our understanding that it should complete quickly enough to call synchronously from the network thread.

Total number of adapters enabled for DHCP as seen when the wait timer in the DHCP WPAD code hits. This timer fires after a timeout from when we get some information from the first adapter to finish.

Number of adapters enabled for DHCP that we have not completed retrieving information for, as seen when the wait timer in the DHCP WPAD code hits. This timer fires after a timeout from when we get some information from the first adapter to finish.

Counts the number of errors from the DhcpRequestParams API that we do not have specific handling for, so that we can see if there is an abnormally high rate.

Deprecated- see Net.DNS_Resolution_And_TCP_Connection_Latency2 Deprecated- see Net.DNS_Resolution_And_TCP_Connection_Latency2

The time measured before starting DNS lookup until after the connection is complete.

Deprecated- see Net.DNS_Resolution_And_TCP_Connection_Latency2

The number of HTTP request responses with MS Office Docs MIME types. The responses are classified based on their method type and cacheability (POST, cacheable GET and non-cacheable GET). The histogram is used in Double GET Experiment, where successful non-cacheable GET requests are intercepted after initial response and repeated in order to determine how much reissuing non-cacheable GET requests influences their error rate. The histogram tracks only initial requests (not the repeated ones).

The response codes encountered for GET request repeated in Double GET Experiment. In the experiment successful non-cacheable GET requests are intercepted after initial response and repeated. The goal of the experiment is to measure how much reissuing non-cacheable GET requests influences their error rate.

Kbps on download streams exceeding 25KB. Measures from the beginning of the first byte received until the end of flowing data.

Net error codes that requests for images end with, including net::OK and net:ERR_ABORTED.

Deprecated as of 2011/5/24, replaced by Net.ErrorCodesForMainFrame2, which measures the same data but uses a different bucket structure (adds guard buckets).

Positive net error code that a page failed with. Note that this only counts the errors in "main frames", so it is a measure of the error pages that users actually see (it does not for example count the error codes for subresoures on a page).

Deprecated as of 2012/5/16, replaced by Net.ErrorCodesForMainFrame3, which measures the same data but includes ERR_ABORTED and OK.

Positive net error codes that requests for pages end with, including net::OK and net::ERR_ABORTED. This only counts loads in "main frames" (it does not for example count the error codes for subresoures on a page).

Deprecated as of 2012/5/16, replaced by Net.ErrorCodesForSubresources2, which measures the same data but includes ERR_ABORT and OK.

Positive net error code that a page failed with. Note that this only counts the errors in "subresources".

Net error codes that requests for "subresources" end with, including net::OK and net::ERR_ABORTED.

System error code that a file Flush failed with. The code is OS dependent, so when looking at the histogram don't mix OSes.

System error code that a file GetSize failed with. The code is OS dependent, so when looking at the histogram don't mix OSes.

System error code that a file Open failed with. The code is OS dependent, so when looking at the histogram don't mix OSes.

System error code that a file Read failed with. The code is OS dependent, so when looking at the histogram don't mix OSes.

System error code that a file Seek failed with. The code is OS dependent, so when looking at the histogram don't mix OSes.

System error code that a file SetEof failed with. The code is OS dependent, so when looking at the histogram don't mix OSes.

System error code that a file Write failed with. The code is OS dependent, so when looking at the histogram don't mix OSes.

System error code range that a file Flush failed with. Any value other than 0 indicates that we have received errors in a range outside of the one in which we recorded the specific errors in Net.FileError_Flush. The code is OS dependent, so when looking at the histogram don't mix OSes.

System error code range that a file GetSize failed with. Any value other than 0 indicates that we have received errors in a range outside of the one in which we recorded the specific errors in Net.FileError_GetSize. The code is OS dependent, so when looking at the histogram don't mix OSes.

System error code range that a file Open failed with. Any value other than 0 indicates that we have received errors in a range outside of the one in which we recorded the specific errors in Net.FileError_Open. The code is OS dependent, so when looking at the histogram don't mix OSes.

System error code range that a file Read failed with. Any value other than 0 indicates that we have received errors in a range outside of the one in which we recorded the specific errors in Net.FileError_Read. The code is OS dependent, so when looking at the histogram don't mix OSes.

System error code range that a file Seek failed with. Any value other than 0 indicates that we have received errors in a range outside of the one in which we recorded the specific errors in Net.FileError_Seek. The code is OS dependent, so when looking at the histogram don't mix OSes.

System error code range that a file SetEof failed with. Any value other than 0 indicates that we have received errors in a range outside of the one in which we recorded the specific errors in Net.FileError_SetEof. The code is OS dependent, so when looking at the histogram don't mix OSes.

System error code range that a file Write failed with. Any value other than 0 indicates that we have received errors in a range outside of the one in which we recorded the specific errors in Net.FileError_Write. The code is OS dependent, so when looking at the histogram don't mix OSes.

The number of times each FTP Error was observed.

The number of Chrome sessions which encountered the indicates FTP Error. This prevents allowing a user that retried a connection many times (getting an error each time) from biasing the tallies.

Each bucket is the number of FTP server types the user has encountered during the session.

The time spent waiting for WinHttpGetProxyForUrl to return with error.

The time spent waiting for WinHttpGetProxyForUrl to return with success.

Nonzero if the HTTP request was to a Google server which required SSL version fallback. The value indicates the SSL version the request fell back on. Since Google servers support TLS 1.2, any fallback is an indication of network middleware problems.

The count was inaccurate (it counted transactions rather than connections).

Each bucket is a boolean (0 or 1) indicating whether the user has had a connection of that type during the session.

This statistic measures successful and failed connections, the new one only measures successful ones.

Each bucket is a boolean (0 or 1) indicating whether the user has had a connection of that type during the session.

Each bucket is a boolean (0 or 1) indicating whether the user has had a successful connection of that type during the session.

Each bucket is a boolean (0 or 1) indicating whether the user has had a connection with an FTP server of that type during the session.

Per-authentication-scheme counts of authentication attempts and rejections.

Count of authentication requests for top level pages vs. sub-resources, such as images or iframes.

Per-authentication-scheme counts of authentication targets, such as secure servers or proxies.

Time between the HttpNetworkTransaction requesting a connection and the time it connected.

Length of time that a received resource will be cacheable.

Size of the response body. This is the actual number of bytes received, which usually agrees with but is not necessarily the same as the size specified by the Content-Length header.

Size of the response body if it is cacheable. This is the actual number of bytes received, which usually agrees with but is not necessarily the same as the size specified by the Content-Length header.

Size of the response body if it is cacheable for at least 24 hours. This is the actual number of bytes received, which usually agrees with but is not necessarily the same as the size specified by the Content-Length header.

Size of the response body if it is cacheable for at least 4 hours. This is the actual number of bytes received, which usually agrees with but is not necessarily the same as the size specified by the Content-Length header.

The difference between the size specified in the X-Original-Content-Length header and the size of teh response body. This is zero if the X-Original-Content-Length header is not present in the response.

The difference between the size specified in the X-Original-Content-Length header and the size of the response body. Only includes resources that have the X-Original-Content-Length header.

Size of the response body. Only includes resources that have the X-Original-Content-Length header.

Time it takes to complete an HttpJob, from starting the transaction until we are done reading.

Time it takes to complete an HttpJob, from starting the transaction until we are done reading, for jobs served from the cache.

Time it takes to complete an HttpJob, from starting the transaction until the job is killed. Note that we didn't detect the end of the data for this job.

Time it takes to complete an HttpJob, from starting the transaction until we are done reading, for jobs not served from the cache.

Time it takes to complete an HttpJob, from starting the transaction until we are done reading, for jobs when we read until no more data is available.

Size specified in the X-Original-Content-Length header. If this header is not present in the response, the size of the response body is used.

Size specified in the X-Original-Content-Length header. Only includes resources that have the X-Original-Content-Length header.

Time it takes to request a new (unused) HTTP proxy socket.

The count of HTTP Response codes encountered.

The count of HTTP Response codes encountered, in response to MAIN_FRAME requests only; saving only the hundreds digit, e.g. 100->1, 300->3.

The counts of the type of sockets (all HTTP sockets, regardless of any proxy used) used for HTTP[s].

Time from when an HTTP request is issued to when the first byte is processed.

The count of handleable socket errors (connection abort/close/reset) per socket reuse type.

Late bindings are on by default now.

The count of handleable socket errors (connection abort/close/reset) per socket reuse type. Socket late binding is disabled.

Late bindings are on by default now.

The count of handleable socket errors (connection abort/close/reset) per socket reuse type. Socket late binding is enabled.

Duration of time spent during the UDP-connect IPv6 probe.

Whether the the interface-enumeration IPv6 probe method failed given that the UDP-connect IPV6 probe failed.

Whether the the interface-enumeration IPv6 probe method was successful given that the UDP-connect IPV6 probe was successful.

The probe results when a test for IPv6 support is done.

The probe results when a test for IPv6 support is done, after a network change event.

The time that a (non-cancelled) proxy resolution request was stalled waiting for an execution thread, for MultiThreadedProxyResolver.

The total time that it took for a (non-cancelled) proxy resolution request to complete, for MultiThreadedProxyResolver.

How often automatically retrying to download the main frame of a page in response to specific HTTP network errors succeeds.

How often automatically retrying to download a subresource in response to specific HTTP network errors succeeds.

How often automatically retrying to download the main frame of a page in response to specific HTTP network errors returns another network error. Histogram includes only the error code that triggered the retry.

How often automatically retrying to download a subresource in response to specific HTTP network errors returns another network error. Histogram includes only the error code that triggered the retry.

On Windows, NetworkChangeNotifierWin calls NotifyAddrChange, which can fail for unknown reasons. This records the number of times it fails in a row before a successful call. If it never succeeds, or takes over 100 tries, a value of 100 is recorded. See http://crbug.com/69198

The number of duplicate cookies that were present in the cookie store during startup.

When validating an HTTPS certificate we may have to make one or more HTTP fetches to OCSP responders in order to get revocation information. This measures the amount of time that failures to get OCSP information take.

When validating an HTTPS certificate we may have to make one or more HTTP fetches to OCSP responders in order to get revocation information. This records the fraction of successful requests.

When validating an HTTPS certificate we may have to make one or more HTTP fetches to OCSP responders in order to get revocation information. This measures the amount of time that each of those requests takes.

Positive error code that was returned by the system library "getaddrinfo()". This error code is platform specific, so when there is a Windows/Linux conflict, both decodings are shown.

Positive error code that was returned by the system library "getaddrinfo()".

Indicate whether a link navigation was preceded by a recent pre-connect trigger (within 10 seconds). There is a high chance that loading the page used a preconnected TCP session.

Indicate whether a URLRequest was preceded by a recent pre-connect trigger (within 10 seconds). There is a high chance that loading the resource used a preconnected TCP session.

When a preconnection is made, indicate what the motivation was.

Currently, the most common (only?) motivations are SELF_REFERAL, LEARNED_REFERAL and OMNIBOX. The SELF_REFERAL indicates that we made sure a second connection was available for a resource that either was never before seen, or has historically had no subresources. The LEARNED_REFERAL indicates that we "learned" that a subresource was commonly needed, and that motivated the TCP/IP preconnect. The OMNIBOX motivation happens when a search is being suggested, and we preconnect to the search provider. (WARNING: Prior to version 7.517.*, enums 7, 8, and 9 may be confused, as EARLY_LOAD_MOTIVATED was inserted new 6 value.)

Indicate whether there was a proxy to preclude preconnection.

What did we decide to do about a predicted resource, based on the historical expected number of connection that this subresource will require.

This is basically the current thresholding of the SubresourceExpectation, relative to current static thresholds, and taking into account whether preconnection is enabled (i.e., if preconnection is disabled, we'll never decide to preconnect).

The expected number of connections, times 100, that we'll make to a given subresource, based on learned history.

By comparing this to thresholds, we decide if we will preconnect, preresolve, or do nothing. This histogram can be used to select those static thresholds.

Indicate whether if a preconnect trigger is followed by a resource request (from link navigations) to the host or not. This is to measure precision of link-based preconnect triggers.

Sourced data corrected, and replaced by NetPreconnectUtilization2

Indicate final utilization for each attempted socket connection.

We also include stats for non-speculative sockets. Some socket connections may never connect, and others may never be used (as the user may abort before then).

Indicate final utilization for each attempted socket connection.

We also include stats for non-speculative sockets. Some socket connections may never connect, and others may never be used (as the user may abort before then).

The completion status of prefetches that have finished loading.

Measurement occurs at ResourceLoader::ResponseCompleted so requests canceled before that point are not registered. Note that "success from cache" means that the UrlRequest::was_cached() was true, and "success from network" means that was_cached() was false. Validated results are considered cached, even though a conditional network request is made.

Number of bytes read from the network on behalf of prefetch requests. This is prefilter, so before any decompression.

Time spent on prefetch requests before the request was canceled.

Time spent on prefetch requests when fetched from cache.

Time spent on prefetch requests when fetched from the network, including validation time.

Replaced by Net.Priority_High_Latency_b.

Time from the start of the http transaction until the first byte of the response for high priority (currently frame and subframe) requests. Only times under 10 minutes are recorded.

Time from the start of the http transaction until the first byte of the response for high priority (currently frame and subframe) requests.

Replaced by Net.Priority_Low_Latency_b.

Time from the start of the http transaction until the first byte of the response for low priority (non-frame/subframe) requests. Only times under 10 minutes are recorded.

Time from the start of the http transaction until the first byte of the response for low priority (non-frame/subframe) requests.

The time in milliseconds spent fetch the system proxy configuration, when polling it for changes.

The total amount of time that was spent executing the proxy script during "tracing" runs (executions of the script which discovered a new DNS dependency and were subsequently abandoned).

The total amount of time that was spent in the non-blocking DNS bindings while executing PAC scripts. This includes the times for abandoned executions.

The amount of time inside of V8 that the proxy script spent executing for the final pass. This includes the time spent in the javascript bindings. This does not include the time spent in abandoned execution passes.

The number of times that alert() was called in the final execution of the script.

The number of errors that were seen in the final execution of the script.

The number of times that the PAC script execution was restarted.

The total time that the proxy resolution took. This includes all the time spent waiting for DNS, PAC script execution, and restarts.

The total time that proxy resolution spent waiting for DNS. This also includes any queuing delays on the origin thread waiting for the DNS result to be processed.

The number of unique DNS hostnames that the PAC script tried to resolve. The *Ex() versions of the bindings count separately.

The total amount of time that was spent in the non-blocking DNS bindings while executing PAC scripts. This includes the times for abandoned executions.

The amount of time inside of V8 that the proxy script spent executing for the final pass. This includes the time spent in the javascript bindings (which is probably dominated by Net.ProxyResolver.DnsWaitTotalTime). This does not include the time spent in abandoned execution passes.

The number of times that alert() was called in the final execution of the script.

The number of errors that were seen in the final execution of the script.

The number of times that the PAC script execution was restarted.

The amount of time it took upon completion to run the final task posted back to the IO thread.

The total time that the proxy resolution took. This includes all the time spent waiting for DNS, PAC script execution, and restarts.

The total time that proxy resolution spent waiting for DNS. This also includes any queuing delays on the origin thread waiting for the DNS result to be processed.

The total time that the proxy resolution took, not including the post back to the origin thread. This includes all the time spent waiting for DNS, PAC script execution, and restarts.

The number of unique DNS hostnames that the PAC script tried to resolve. The *Ex() versions of the bindings count separately.

The length of the URL that was passed into the PAC script.

Second-level domains for which we have observed public key pinning failures.

A validated certificate chain may be subject to additional "pinning" requirements on a per-domain basis. This records the fraction of successful matches between a certificate chain and a pin list.

The state of a QUIC connection's crypto hanshake as it progresses from starting to confirmation or failure.

The number of ports suggested per server.

The number of client hello messages sent.

The number of client hello messages sent when the crypto handshake was confirmed.

The number of stream frames bundled within a received packet.

The number of stream frames per stream ID within a received packet.

The network error code which resulted in the session being closed.

The number of streams open when a QUIC session timed out.

The number of total streams created when a QUIC session crypto handshake timed out.

The QUIC error code which resulted in the connection being closed.

The QUIC error code which resulted in the connection being closed by the client.

The QUIC error code which resulted in the connection being closed by the server.

The value of the TCP cubic sender's CWND when the session is closed.

The number of QUIC streams opened when a new QUIC stream is created.

The total number of streams created by the client when the session is closed.

The number of missing packets between the current received packet and the previously largest received packet sequence number, when the current received packet had a lower sequence number than the previously received packet sequence number.

The number of times the current received packet had a lower sequence number than the previously received packet sequence number.

The number of missing packets between the current received packet and the previously largest received packet sequence number.

The number of missing packets between the current received packet and the previously largest received packet sequence number, as reported by the remote end of the connection.

Version of the QUIC protocol used for this connection.

The network error code returned when attempting to read to a QUIC connection.

The QUIC error code which resulted in a stream being reset by the client.

The QUIC error code which resulted in a stream being reset by the server.

The network error code returned when attempting to write to a QUIC connection.

True if the HTTP request was sent to a server which supports the TLS renegotiation extension.

When starting a cross-site navigation, the time between reading the headers and body of the response.

The time an already used socket sat idle before being used.

The time an unused socket (all HTTP sockets, regardless of any proxy used) sat idle before being used.

The time a previously used socket sat idle before encountering a recoverable socket IO error (connection abort/reset/close).

The time an unused socket sat idle before encountering a recoverable socket IO error (connection abort/reset/close).

Net error codes that socket initializations end with, including net::OK and net::ERR_ABORTED.

Time in milliseconds from initial RequestSocket() call until successfully acquiring a connected socket.

The time from the connection start to connection establish.

The time waiting to be ready to start connecting.

Each bucket is the number of connection type of socket stream.

The time a socket stream was open.

Each bucket is the number of protocol type on socket stream.

Number of bytes on a socket stream.

Number of reads on a socket stream.

Number of bytes on a socket stream.

Number of Write on a socket stream.

The counts of the type of sockets returned by the socket pools.

see SocketIdleTimeBeforeNextUse_ReusedSocket_SOCK

The time an already used SOCKS socket sat idle before being used.

see SocketIdleTimeBeforeNextUse_UnusedSocket_SOCK

The time an unused SOCKS socket sat idle before being used.

see SocketRequestTime_SOCK

Time from initial SOCKSClientSocketPool::RequestSocket() call until successfully acquiring a connected SOCKS socket.

Time it takes to request a new (unused) SOCKS proxy socket.

see SocketType_SOCK

The counts of the type of sockets returned by the SOCKS pool.

Time from when the Connect() starts until it completes.

The counts of the flow control state of each frame (with stream and session flow control on).

The counts of the flow control state of each frame (with stream flow control on).

Status of checking if a SPDY domain can handle a IP match. If a match is found, we successfully used the IP Pooling. If a match is not found, we could have used IP Pooling, except the TLS Cert didn't match the IP-pooled domain.

The RTT for SPDY's PING.

The count of streams at each priority over Spdy sessions.

The number of bytes recevied per stream.

The number of bytes sent per stream.

Total number of bytes recevied per session before closing session due to EOF.

Total number of bytes recevied per session before closing session due to an error during read.

Net error codes when SpdySession was closed, doesn't inlcuding net::OK.

Socket connected status in SpdySession::CreateStream.

Replaced by SpdySessionErrorDetails2 on 2013-04-19.

WARNING: r181910 added an enum value in the middle, so don't trust the counts for values 9 and above for Chrome builds after that revision. The type of SPDY Protocol error encountered.

The type of SPDY Protocol error encountered.

Replaced by SpdySessionErrorDetails_Google2 on 2013-04-19.

The type of SPDY Protocol error encountered when talking to a google.com server.

WARNING: r181910 added an enum value in the middle, so don't trust the counts for values 9 and above for Chrome builds after that revision. The type of SPDY Protocol error encountered when talking to a google.com server.

The type of SPDY Session used when looking up a session.

Whether SpdySession::Get{Peer,Local}Address was called when the connection had no socket.

The count of SPDY sessions using the data reduction proxy and the count of other SPDY sessions.

SpdySession::GetLocalAddress returned ERR_SOCKET_NOT_CONNECTED.

SpdySession::GetPeerAddress returned ERR_SOCKET_NOT_CONNECTED.

The count of SPDY Sessions with or without stalls.

The congestion window (in pkts) received at the end of a SpdySession.

The congestion window (in pkts) sent at the beginning of a SpdySession.

Percentage of sessions which received settings from the server.

The Download Retransmission Rate (%) received at the end of a SpdySession.

The RTT received at the end of a SpdySession.

Percentage of sessions which sent settings to the server.

The time between receiving the the first chunk and the last chunk of data on a Spdy stream.

The number of pushed, but abandoned streams over a single session.

The number of streams issued over a single session.

The number of pushed, and used streams over a single session.

The number of push streams received over a single session.

The number of stream stalls per session.

The time of a Spdy stream. Measured from sending the first chunk to receiving the last chunk of data.

The time between sending the request and receiving the first chunk of data on a Spdy stream.

The percent compression achieved when compression SYN_STREAM frames.

The SPDY protocol version that is used to talk to SPDY servers.

The SSL/TLS cipher suite that was negotiated.

Time from when the Connect() starts until it completes.

Time from when the Connect() starts until it completes when using the data reduction proxy. This includes certificate retrieval and verification.

Time from when the Connect() starts until it completes for google.com and any subdomain of it.

Time from when the Connect() starts until it completes for google.com and any subdomain of it. This only includes users in a 50% field trial that disables revocation checking for certificate pinned sites.

Time from when the Connect() starts until it completes for google.com and any subdomain of it. This only includes users not in a 50% field trail that disables revocation for certificate pinned sites.

Counts the number of times that users have hit blacklisted certificates. The indexes match up to the indexes in net/base/x509_certificate.cc:IsBlacklisted. The details of the certificates in question is confidential.

Time to complete a certificate verification (success case).

Time to complete a certificate verification (error case).

Time to complete a DNS lookup for a DNS CAA record.

Time that we would have wasted had we waited for a CAA lookup in order to validate a certificate.

Time to complete a speculative certificate verification.

The number of times that we have performed SSLv3 fallback and found a TLS renegotiation patched server.

Was a speculative certificate verification used?

Time saved by a speculative certificate vertification.

Number of idle sockets when the Connect() succeeded.

Time from when the Connect() starts until it completes. Only times under 10 minutes are logged.

Time from when the Connect() starts until it completes when the network address only contains IPv4 addresses. Only times under 10 minutes are logged.

Time from when the Connect() starts until it completes when the IPv4 fallback connection won the race against IPv6. Only times under 10 minutes are logged.

Time from when the Connect() starts until it completes when we race an IPv6 connection against an IPv4 connection with a 300ms delay. Only times under 10 minutes are logged.

Time from when the Connect() starts until it completes when the network address only contains IPv6 addresses. Only times under 10 minutes are logged.

For sockets for which a TCP Fast Open protocol might be used, the result of trying to use it.

see SocketIdleTimeBeforeNextUse_ReusedSocket_TCPforSOCKS

The time an already used TCP socket sat idle before being used for a SOCKS request.

see SocketIdleTimeBeforeNextUse_UnusedSocket_TCPforSOCKS

The time an unused TCP socket sat idle before being used for a SOCKS request.

see SocketRequestTime_TCPforSOCKS

Time from initial SOCKSClientSocketPool::RequestSocket() call until successfully acquiring a connected TCP socket.

see SocketType_TCPforSOCKS

The counts of the type of sockets returned by the TCP pool used by the SOCKS pool.

Was only used for HTTP[S] connections, renamed to Net.HTTPSocketType.

The counts of the type of TCP socket returned.

(discontinued as of 4/12/09) Effective bandwidth in KByte/Second of transactions logged to Transaction_Latency histogram. Note that only samples durations greater than zero ms, and less than 1 hour are tallied into this ratio.

Time from the when the network transaction is requested, until the first byte of the header is received.

Replaced by Net.Transaction_Connected_New_b.

When a new connection is established, the time from the when the network transaction is requested, until the first byte of the header is received. Only items under 10 minutes are logged.

When a new connection is established, the time from the when the network transaction is requested, until the first byte of the header is received.

Replaced by Net.Transaction_Connected.

Time from the when the network transaction is requested, until the first byte of the header is received. Only items under 10 minutes are logged.

Replaced by Net.Transaction_Latency_b.

Time from first byte sent until last byte received by the new network stack. Only items under 1 hour are logged.

Time from first byte sent until last byte received by the new network stack.

Time from when a network transaction is requested until last byte received by the new network stack.

When an existing TCP/IP connection is NOT reused, the time from when a network transaction is requested until last byte received by the new network stack.

Replaced by Net.Transaction_Latency_Total_New_Connection.

When an existing TCP/IP connection is NOT reused, the time from when a network transaction is requested until last byte received by the new network stack. Only items under 10 minutes are logged.

Replaced by Net.Transaction_Latency_Total.

Time from when a network transaction is requested until last byte received by the new network stack. Only items under 10 minutes are logged.

Replaced by Net.Transaction_Latency.

Time from first byte sent until last byte received by the new network stack. Only items under 10 minutes are logged.

Time from first byte sent until last byte received with old WinHTTP network stack. Only items under 1 hour are logged.

The time an already used TCP socket sat idle before being used (either for direct or non-socks use).

The time an unused TCP socket sat idle before being used (either for direct or non-socks use).

Time from initial ClientSocketPool::RequestSocket() call until successfully acquiring a connected socket (either for direct or non-socks use).

The counts of the type of sockets returned by the TCP pool (either for direct or non-socks use).

Posix error code from call to bind() UDP socket.

Windows error code from call to bind() UDP socket.

Chromium error code from call to RandomBind() UDP socket.

The time spent in closesocket call in UDPSocketWin::Close.

The number of Wi-fi adapters on the computer. Because the histogram is logged each time Chrome performs a Wi-fi scan, it's better to see results in the "user count" view.

The time that a request to Location Based Services takes.

The time that a Wi-fi scan takes.

Duration of time that a failing WPAD QuickCheck takes. WPAD QuickCheck does a name lookup for "wpad" and times out quickly to fail fast when there's no WPAD server on the network.

Duration of time that a successful WPAD QuickCheck takes. WPAD QuickCheck does a name lookup for "wpad" and times out quickly to fail fast when there's no WPAD server on the network.

The network error, if any, of the first pipeline connectivity request.

The HTTP response code, if any, of the first pipeline connectivity response.

The result of the first pipeline connectivity request.

The network error, if any, of the second pipeline connectivity request.

The HTTP response code, if any, of the second pipeline connectivity response.

The result of the second pipeline connectivity request.

The network error, if any, of the third pipeline connectivity request.

The HTTP response code, if any, of the third pipeline connectivity response.

The result of the third pipeline connectivity request.

The network error, if any, of the fourth pipeline connectivity request.

The HTTP response code, if any, of the fourth pipeline connectivity response.

The result of the fourth pipeline connectivity request.

The network error, if any, of the fifth pipeline connectivity request.

The HTTP response code, if any, of the fifth pipeline connectivity response.

The result of the fifth pipeline connectivity request.

The network error, if any, of the stats pipeline connectivity request.

The HTTP response code, if any, of the stats pipeline connectivity response.

The result of the stats pipeline connectivity request.

True if all requests received by the pipelining test server were HTTP/1.1.

True if the non-pipelined canary request sent immediately before the pipelining test requests succeeded. Note that if this fails, the rest of the NetConnectivity.Pipeline.* stats are not collected.

The maximum depth of pipelined requests received by the test server.

True if the entire pipeline connectivity trial passed.

Deprecated 6/25/2012. No longer tracked.

In this experiment, 21 packets were sent to Google via UDP at port 6121 as rapidly as possible, just after successfully sending an UMA upload. Each packet was numbered, as was its ACK sent back by Google. If no packets (of the 21) were ever ACKed, then the port is assumed to be blocked, and no data is recorded in this histogram. If the port is not blocked, then this histogram shows the number of echo responses received from the first

Deprecated 6/25/2012. No longer tracked.

In this experiment, 21 packets were sent to Google via UDP at port 6121 as rapidly as possible, just after successfully sending an UMA upload. If no packets (of the 21) were ever ACKed, then the port is assumed to be blocked. The histogram shows if we ever got an ACK for a packet in our series of 21.

Deprecated 4/2012. No longer tracked.

The RTT for echoing 100 bytes of TCP data unsuccessfully.

Deprecated 4/2012. No longer tracked.

The RTT for echoing 1K bytes of TCP data successfully.

Status for TCP protocol for echoing

Deprecated 4/2012. No longer tracked.

Status for echoing 100 bytes of TCP data.

Deprecated 4/2012. No longer tracked.

Status for echoing 1K bytes of TCP data.

The RTT for TCP protocol for echoing

Deprecated 4/2012. No longer tracked.

The RTT for echoing 100 bytes of TCP data successfully.

Deprecated 4/2012. No longer tracked.

The RTT for echoing 1K bytes of TCP data successfully.

Deprecated 4/2012. No longer tracked.

The RTT for echoing 100 bytes of UDP data unsuccessfully.

Deprecated 4/2012. No longer tracked.

The RTT for echoing 1K bytes of UDP data successfully.

Deprecated 6/25/2012. No longer tracked.

Chrome sends 4 UDP packets in a row to test to see if there is a probabalistic dependency in packet loss for consecutive packets. We record a bit vector of packets received, where the least significant bit is a 1 if the first packet was received, etc. For example, if packets 1 and 3 are received, but packets 2 and 4 are lost, then we'd record a sample of binary 0101B, or 5.

Deprecated 6/25/2012. No longer tracked.

Chrome sends 6 UDP packets in a row to test to see if there is a probabalistic dependency in packet loss for consecutive packets. We record a bit vector of packets received, where the least significant bit is a 1 if the first packet was received, etc. For example, if all packets other than packet 2 and 4 are responded to, then we'd have a sample (in binary) of 110101B, or 53.

Status for UDP protocol for echoing

Deprecated 4/2012. No longer tracked.

Status for echoing 100 bytes of UDP data.

Deprecated 4/2012. No longer tracked.

Status for echoing 1K bytes of UDP data.

The RTT for UDP protocol for echoing

Deprecated 4/2012. No longer tracked.

The RTT for echoing 100 bytes of UDP data successfully.

Deprecated 4/2012. No longer tracked.

The RTT for echoing 1k bytes of UDP data successfully.

This histogram records how many packets (out of 6 attempted) were sent via UDP as rapidly as possible, just after successfully sending an UMA upload.

In this experiment, 21 packets were sent to Google via UDP as rapidly as possible, just after successfully sending an UMA upload. Each packet was numbered, as was its ACK sent back by Google. If no packets (of the 21) were ever ACKed, then the port is assumed to be blocked, and no data is recorded in this histogram. If the port is not blocked, then this histogram shows the number of echo responses received from the first

In this experiment, 21 packets were sent to Google via UDP as rapidly as possible, just after successfully sending an UMA upload. If no packets (of the 21) were ever ACKed, then the port is assumed to be blocked. The histogram shows if we ever got an ACK for a packet in our series of 21.

This histogram records how many packets (out of 21 attempted) were sent via UDP as rapidly as possible, just after successfully sending an UMA upload.

In this experiment, 21 packets were sent to Google via UDP on port 443 or 6121.

Deprecated 9/2012. No longer tracked.

This histogram records how many packets (out of 6 attempted) were sent via UDP as rapidly as possible, just after successfully sending an UMA upload.

In this experiment, a few packets were sent from Google to clients via UDP on port 443 or 80 to perform net connectivity test.

Next NetConnectivity5 experiment weren't started because there is an outstading pending write.

The time the Gobi modem takes to complete activation.

The time the Gobi modem takes to connect to the cellular network.

The time the Gobi modem takes to disconnect from the cellular network.

Number of attempts taken to install Gobi firmware.

The time it takes to install Gobi firmware.

The time the Gobi modem takes to register on the cellular network.

Errors experienced during Gobi device powerup.

Chrome OS network metric sampling the time spent using Cellular to transport data. These data are mostly useful when summed and compared to TimeOnline for other network technologies (e.g. WiFi vs Cellular).

Chrome OS network performance metric sampling the time to join a 3G/Cellular network and configure Layer 3 state.

Chrome OS network performance metric sampling the time to determine that a 3G/Cellular network is online after configuring Layer 3 state.

Chrome OS network performance metric sampling the time to determine that a 3G/Cellular network is in a captive portal after configuring Layer 3 state.

Chrome OS cellular usage API request status codes.

Chrome OS network metric sampling the time spent using Ethernet to transport data. These data are mostly useful when summed and compared to TimeOnline for other network technologies (e.g. WiFi vs Cellular).

Chrome OS network performance metric sampling the time to join a wired Ethernet network and configure Layer 3 state (typically acquire a DHCP lease).

Chrome OS network performance metric sampling the time to determine that an Ethernet network is online after configuring Layer 3 state.

Chrome OS network performance metric sampling the time to determine that an Ethernet network is in a captive portal after configuring Layer 3 state.

Chrome OS metric counting the number of network configurations that contained a NSS nickname identifying a CA certificate, which triggered the migration to PEM encoding. This metric doesn't consider whether the migration was successful but once a migration was successful the nickname is removed.

Chrome OS connection manager service errors seen.

Chrome OS network diagnostic metric sampling the number of cellular network flakes. A network flake occurs when the signal strength goes below detection level for a short duration.

Chrome OS network diagnostic metric sampling the total amount of time spent from the start of the first auto-connect request until when the cellular modem successfully connects to the network.

Chrome OS network diagnostic metric sampling the number of auto-connect tries that were attempted before the cellular modem successfully connected to the network.

Chrome OS network metric that tracks the number of DHCP option failures encountered by Shill. This indicates that Shill is using minimal DHCP options due to suspected MTU issues on the return path from the DHCP server back to the client.

Chrome OS network usage metric that tracks whether the cellular network was disconnected due to an error or was explicitly disconnected by the user.

Chrome OS cellular network metric that tracks the number of drops based on the network technology.

Chrome OS cellular network metric that tracks the number of out-of-credits detected based on the cause that triggered the out-of-credits.

Chrome OS network diagnostic metric sampling the number of portal detection attempts per pass for a cellular network. This includes failure, timeout and successful attempts.

Chrome OS network diagnostic metric sampling the total number of portal detection attempts performed for a cellular network between the Connected and Online state. This includes failure, timeout and successful attempts.

Chrome OS network diagnostic metric sampling the result of portal detections for a cellular network.

Chrome OS network metric sampling the signal strength (0-100) of the cellular modem before it dropped from the network.

Chrome OS network metric sampling the time spent using cellular to transport data. These data are mostly useful when summed and compared to TimeOnline for other network technologies (e.g. WiFi vs Cellular).

Chrome OS network performance metric sampling the time to join a cellular network and configure Layer 3 state.

Chrome OS network performance metric sampling the time to connect a cellular modem.

Chrome OS network performance metric sampling the time to disable a cellular modem.

Chrome OS network performance metric sampling the time to enable a cellular modem.

Chrome OS network performance metric sampling the time to initialize a cellular modem.

Chrome OS network performance metric sampling the time to determine that a cellular network is online after configuring Layer 3 state.

Chrome OS network performance metric sampling the time to determine that a cellular network is in a captive portal after configuring Layer 3 state.

Chrome OS network performance metric sampling the time to scan a cellular network and register a modem.

Chrome OS cellular network metric that tracks the number of corrupted profiles encountered by Shill.

Chrome OS network usage metric that tracks whether the Ethernet network was disconnected due to an error or was explicitly disconnected by the user.

Chrome OS network performance metric that tracks the number of LinkMonitor broadcast errors that were accrued on an Ethernet network at the time that the link was declaired to be failed.

Chrome OS metric that signals the type of failure the LinkMonitor encountered which caused it to stop monitoring an Ethernet network.

Chrome OS network performance metric that tracks the number of milliseconds between an ARP request and a received reply on an Ethernet network.

Chrome OS network performance metric that tracks the number of seconds from the start of the LinkMonitor until failure on an Ethernet network.

Chrome OS network performance metric that tracks the number of LinkMonitor unicast errors that were accrued on an Ethernet network at the time that the link was declaired to be failed.

Chrome OS network diagnostic metric sampling the number of portal detection attempts per pass for an Ethernet network. This includes failure, timeout and successful attempts.

Chrome OS network diagnostic metric sampling the total number of portal detection attempts performed for an Ethernet network between the Connected and Online state. This includes failure, timeout and successful attempts.

Chrome OS network diagnostic metric sampling the result of portal detections for an Ethernet network.

Chrome OS network performance metric sampling the time to join a wired Ethernet network and configure Layer 3 state (typically acquire a DHCP lease).

Chrome OS network performance metric sampling the time to initialize an Ethernet device.

Chrome OS network performance metric sampling the time to determine that an Ethernet network is online after configuring Layer 3 state.

Chrome OS network performance metric sampling the time to determine that an Ethernet network is in a captive portal after configuring Layer 3 state.

Chrome OS connection manager service errors seen.

Deprecated 10/2012. No longer tracked.

Chrome OS network diagnostic metric sampling the number of termination actions that successfully complete or fail when shill terminates.

Chrome OS network diagnostic metric sampling the number of termination actions that successfully complete or fail when shill suspends.

Chrome OS network diagnostic metric sampling the number of termination actions that successfully complete or fail when shill terminates.

Chrome OS network diagnostic metric sampling the time in milliseconds it takes termination actions to complete when shill suspends.

Chrome OS network diagnostic metric sampling the time in milliseconds it takes termination actions to complete when shill terminates.

Chrome OS network stability metric sampling the time in seconds between the networking going online to going offline. Offline events due to device shutdown or suspend are ignored (along with the online time before that offline event).

Chrome OS network usage metric sampled on each successful VPN connection that tracks the VPN connection type.

Chrome OS network usage metric sampled on each successful VPN connection that tracks the remote authentication method.

Chrome OS network performance metric sampling the time to configure Layer 3 state on a VPN network (typically acquire a DHCP lease).

Chrome OS network performance metric sampling the time to determine that a WiMax network is online after configuring Layer 3 state.

Chrome OS network usage metric sampled on each successful VPN connection that tracks the user authentication method.

Chrome OS network usage metric. Reason code reported when the AP disconnects a wifi connection.

Chrome OS network usage metric. Broad category of reason AP disconnected a WiFi connection.

Chrome OS network usage metric. The AP mode setting for each successful WiFi connection.

Chrome OS network usage metric. The channel used for each successful WiFi connection.

Chrome OS network usage metric. Reason code reported when the client disconnects a wifi connection.

Chrome OS network usage metric. Broad category of reason client disconnected a WiFi connection.

Chrome OS network usage metric that tracks whether an 802.11 wireless network was disconnected due to an error or was explicitly disconnected by the user.

Chrome OS network usage metric sampled on each successful 802.1x wireless connection that tracks the configured inner authentication method.

Chrome OS network usage metric sampled on each successful 802.1x wireless connection that tracks the configured outer authentication method.

Chrome OS metric sampling the number of different frequencies (i.e. channels) on which a device has connected to a WiFi network. This value is sampled every time a WiFi connection is established (WPASupplicant::kInterfaceStateCompleted). Note that the word "Ever" in the metric name is misleading. ChromeOS actually ages out historical information, currently after 3 weeks.

Chrome OS network performance metric that tracks the number of LinkMonitor broadcast errors that were accrued on an 802.11 wireiless network at the time that the link was declaired to be failed.

Chrome OS metric that signals the type of failure the LinkMonitor encountered which caused it to stop monitoring an 802.11 wireless network.

Chrome OS network performance metric that tracks the number of milliseconds between an ARP request and a received reply on an 802.11 wireless network.

Chrome OS network performance metric that tracks the number of seconds from the start of the LinkMonitor until failure on an 802.11 wireless network.

Chrome OS network performance metric that tracks the number of LinkMonitor unicast errors that were accrued on an 802.11 wireless network at the time that the link was declaired to be failed.

Chrome OS network usage metric. The channel type used for each successful WiFi connection.

Chrome OS network diagnostic metric sampling the number of portal detection attempts per pass for an 802.11 wireless network. This includes failure, timeout and successful attempts.

Chrome OS network diagnostic metric sampling the total number of portal detection attempts performed for an 802.11 wireless network between the Connected and Online state. This includes failure, timeout and successful attempts.

Chrome OS network diagnostic metric sampling the result of portal detections for an 802.11 wireless network.

Chrome OS network usage metric describing, for a wifi scan attempt, what scan method is used and whether it ends in a connection.

Chrome OS network usage metric describing, for a wifi scan attempt, how many milliseconds were spent waiting to talk to the kernel/drivers.

Chrome OS network usage metric. The security setting for each successful WiFi connection.

Chrome OS network metric indicating the negative of the dBm received signal strength recorded at the time a successful WiFi connection started.

Chrome OS network metric sampling the time spent using WiFi to transport data. These data are mostly useful when summed and compared to TimeOnline for other network technologies (e.g. WiFi vs Cellular).

Chrome OS network performance metric sampling the time from the resume event to the time when an 802.11 wireless network has configured its Layer 3 state.

Chrome OS network performance metric sampling the time to configure Layer 3 state on an 802.11 wireless network (typically acquire a DHCP lease).

Chrome OS network performance metric sampling the time to connect to a WiFi Basic Service Set (which consists of the access point and associated stations on a particular WiFi channel for a specific network).

Chrome OS network performance metric sampling the time to initialize an 802.11 wireless device.

Chrome OS network performance metric sampling the time to join (associate plus authenticate) an 802.11 wireless network.

Chrome OS network performance metric sampling the time to determine that an 802.11 wireless network is online after configuring Layer 3 state.

Chrome OS network performance metric sampling the time to determine that an 802.11 wireless network is in a captive portal after configuring Layer 3 state.

Chrome OS network performance metric sampling the time to scan WiFi until a connection is found.

Chrome OS network performance metric sampling the time between the beginning of a WiFi scan (if the scan includes both a progressive scan and a full scan, the TimeToScanAndConnect starts with the first scan of the series) and the completion of a successful connection.

Chrome OS network performance metric sampling the time to configure Layer 3 state on a WiMax network (typically acquire a DHCP lease).

Chrome OS network performance metric sampling the time to initialize a WiMax device.

Chrome OS network performance metric sampling the time to determine that a WiMax network is online after configuring Layer 3 state.

Chrome OS network performance metric sampling the time to join a 3G/Cellular network and configure Layer 3 state. Note this metric is deprecated; see Network.Cellular.TimeToConfig.

Chrome OS network performance metric sampling the time to join a wired Ethernet network and configure Layer 3 state (typically acquire a DHCP lease). Note this metric is deprecated; see Network.Ethernet.TimeToConfig.

Chrome OS network performance metric sampling the time to configure Layer 3 state on an 802.11 wireless network (typically acquire a DHCP lease). Note this metric is deprecated; see Network.Wifi.TimeToConfig.

Chrome OS network performance metric sampling the time to join (associate plus authenticate) an 802.11 wireless network. Note this metric is deprecated; see Network.Wifi.TimeToJoin.

Chrome OS network performance metric sampling the time to configure Layer 3 state on an 802.11 wireless network (typically acquire a DHCP lease).

Network metric reporting the download speed test results run at setup time. Recorded at least once per day.

Chrome OS network usage metric. The channel used for each successful WiFi connection.

Network metric indicating the negative of the dBm noise level recorded at the time the metric is collected. Reported at least once per day and only when the device is idle.

Network metric indicating the negative of the dBm received signal level recorded at the time the metric is collected. Reported at least once per day and only when the device is idle.

Network metric indicating signal minus noise in dBm recorded at the time the metrics is collected. Reported at least once per day and only when the device is idle.

Network metric indicating the negative of the dBm noise level recorded at the time the metric is collected. Reported at least once per day.

Chrome OS network usage metric. The channel type used for each successful WiFi connection.

Network metric reporting the average round trip time to the wifi gateway. Recorded at least once per day.

Chrome OS network usage metric. The security setting for each successful WiFi connection.

Network metric indicating the negative of the dBm received signal level recorded at the time the metric is collected. Reported at least once per day.

Network metric indicating signal minus noise in dBm recorded at the time the metrics is collected. Reported at least once per day.

Chrome OS network performance metric sampling the time from the resume event to the time when an 802.11 wireless network has configured its Layer 3 state.

Chrome OS network performance metric sampling the time to configure Layer 3 state on an 802.11 wireless network (typically acquire a DHCP lease).

Chrome OS network performance metric sampling the time to join (associate plus authenticate) an 802.11 wireless network.

Chrome OS network performance metric sampling the time to determine that an 802.11 wireless network is online after configuring Layer 3 state.

Chrome OS network performance metric sampling the time to determine that an 802.11 wireless network is in a captive portal after configuring Layer 3 state.

Actions taken by users from the new tab page on Android. These actions may navigate away from the NTP (e.g. searching in the omnibox or opening a bookmark), but can also happen without navigating away from the NTP (e.g. opening a bookmark in a new tab).

Histogram for the source of app page drags. For any succesful drop onto an apps pane of the NTP, this logs where the drag originated.

Actions taken by users on partner bookmarks (editing / renaming) on the NTP on Android.

The default pane when the NTP is first opened.

Histogram of the time, in milliseconds, users have the cursor over a most visited thumbnail before clicking.

Histogram of the time, in milliseconds, users have the cursor over a most visited thumbnail before moving it away from the thumbnail without clicking.

Android: Tallies counts for how the user interacted with the NTP promo page.

Histogram for user clicks of the most visited thumbnails. The value is equal to the index of the thumbnail.

Action taken by the user on the Most Visited NTP pane. If the user switches panes during this use of the NTP, this action is sometimes not recorded. Ask mpearson@ for details.

Records anomalous events for the Most Visited Tile Placement experiment, where it is unable to operate as expected. These are recorded during New Tab Page load time, once for every NTP.

The number of screenshots that were cached for the non-visible but ranked suggestions on the Suggested NTP pane.

Given that the user has typed a URL, and given that that specific URL was ranked but not visible on the Suggested pane of the NTP, this is the rank that the Suggested pane had for that URL.

The number of external tiles that are displayed on the NTP. External tiles are those for which the visuals are handled by the page itself, not by the iframe. Recorded before changing focus away from the NTP, be it by navigating to a URL, switching tabs, changing the active window or closing the tab/shutting down Chrome.

The total number of times the user hovered the mouse over Most Visited tile or title elements before changing focus away from the NTP, be it by navigating to a URL, switching tabs, changing the active window or closing the tab/shutting down Chrome.

Histogram for usage of the menu on the NTP that allows the user to access tabs from other devices.

The pane that had been previously selected when the user switches panes in the NTP.

Histogram for NTP bubble promo activity.

Histogram for NTP notification promo activity.

TBD.

The pane selected when the user switches panes in the NTP.

Histogram for user clicks of the Recently Closed items. The value is the recency of the entry being restored (0 is most recent).

Histogram to track how many times a user switched pages in a single NTP session.

Histogram for user clicks of the suggested site thumbnails. The value is equal to the index of the thumbnail.

Action taken by the user on the Suggested Sites NTP pane.

Time to load the Suggested Sites NTP pane, in milliseconds.

Time spent on the Suggested Sites NTP pane, in seconds.

Histogram for impressions on the various most visited tiles. The value is equal to the index of the thumbnail.

Indicate, for each impression of the New Tab Page, whether the suggestions were obtained from the client or server. Recorded before changing focus away from the NTP, be it by navigating to a URL, switching tabs, changing the active window or closing the tab/shutting down Chrome.

The percentage of errors per attempts to load image thumbnails on the New Tab Page. When an error occurs, a grey tile is shown instead of a thumbnail image. We measure the rate instead of the number of errors because multiple attempts are made to load images at different times during the NTP's lifetime. Each NTP session's error rate is logged after the user navigates to a new URL from that NTP.

The percentage of times most visited tiles use the fallback thumbnail. Only requests that actually specify a fallback thumbnail are considered here. We measure the rate instead of the number of errors because multiple attempts are made to load thumbnails at different times during the NTP's lifetime. Each NTP session's error rate is logged after the user navigates to a new URL from that NTP.

The number of screenshots that were cached for the visible suggestions on the Suggested NTP pane.

Given that the user has typed a URL, and given that that specific URL was visible on the Suggested pane of the NTP, this is the rank that the Suggested pane had for that URL.

The actions taken on notifications, recorded every time they happen. This histogram will record every single event that happens separately.

The actions taken on notifications, recorded once per notification, when it is closed. This differs from the Notifications.Actions histogram in that multiple events of the same type on a single notification will only record a single UMA event.

TBD

Deprecated 10/2011. No longer tracked, replaced with NewTabPage.DefaultPageType

The default pane when the NTP is first opened.

Deprecated 10/2011. No longer tracked, replaced with NewTabPage.SelectedPageType

The pane selected when the user switches panes in the NTP.

Failure reason of final OAuth2 access token retreival call during ChromeOS login.

Retry reason of failed OAuth2 access token retreival call call during ChromeOS login.

Failure reason of final ListAccounts call failure during ChromeOS login.

Retry reason of failed ListAccounts call during ChromeOS login.

Failure reason of final MergeSession call during ChromeOS login.

Retry reason of failed MergeSession call during ChromeOS login.

Failure reason of final OAuthLogin (with SID+LSID) call during ChromeOS login.

Retry reason of failed OAuthLogin (with SID+LSID) call during ChromeOS login.

Failure reason of final OAuthLogin (with uber token) call during ChromeOS login.

Retry reason of failed OAuthLogin (with uber token) call during ChromeOS login.

Outcome of ChromeOS GAIA cookie post merge session verification process. It measures how often /MergeSession request collided with browser session restore process resulting in partially authenticated primary GAIA session.

Outcome of ChromeOS GAIA cookie session restore process.

How long it takes for the session restore to fail.

How long it takes for the session restore to finish succeessfully.

When a page is loaded in offline mode, the percentage of resources on that page that were successfully loaded.

Aggressive HistoryURL provider field trial deleted in spring 2012.

A number that indicates what omnibox ranking behavior the user is seeing as part of the OmniboxAggressiveHistoryURLProvider field trial (OmniboxAggressiveHistoryURLProvider).

The number of cut or copy commands on all selected text in the omnibox. Gathered on desktop platforms (Win, Mac, Linux, ChromeOS).

The number of times users enter keyword hint mode "Search ___ for:" and how.

The length of time between when a user focused on the omnibox and first modifies the omnibox.

The length of time between when a user focused on the omnibox and opened an omnibox match (which could be what they typed or a suggestion).

The number of paste commands on the text in the omnibox. Reported every time a paste command is done.

The number of paste-and-go commands on the text in the omnibox. Reported every time a paste-and-go command is done.

The length of time taken by the named provider"s synchronous pass.

Deprecated 2012-11-14. Replaced by Autocomplete.BookmarkProviderMatchTime.

Time the HistoryContentProvider takes to perform a bookmark search.

Time it takes for the omnibox to become responsive to user input after the user has typed N characters. This measures the time it takes to start all the asynchronous autocomplete providers (but not wait for them to finish).

The id of search engine that was used for search in omnibox. See src/chrome/browser/search_engines/template_url_prepopulate_data.cc for more info.

The type of search engine associated with a match opened from the omnibox.

The time elapsed between the sending of a suggest request to Google until the time the request was returned with status==failed. Ignores requests that were canceled before being returned.

The time elapsed between the sending of a suggest request to Google until the time the request was returned with status==success. Ignores requests that were canceled before being returned.

Counts about the number of suggest requests the omnibox sent, invalidated, and replies received.

The number of most visited suggestions returned when ZeroSuggest would have triggered. The suggestions appear when the user has focused but not modified the omnibox.

Counts about the number of zero suggest requests (requests for suggestions when the user has focused but not modified the omnibox) the omnibox sent, invalidated, and replies received.

The cat's flavor and how many bits there are in it.

Events seen by the OSX NSException swizzle.

Counts the number of times the user clicked on the later button of the outdated upgrade bubble, before clicking on the reinstall button in the same Chrome session.

Completed overscroll gestures.

An overscroll gesture starts when user scrolls past the edge of the web page and continues scrolling in the same direction. An overscroll gesture is completed when user stops scrolling (e.g. by lifting the fingers from the touchscreen or touchpad).

Navigations that were triggered due to completed overscroll gesture. Note that not all completed overscroll gestures trigger a navigation.

Overscroll gestures initiated by the user. Note that not all overcroll gestures started are completed (e.g. the overscroll gesture is aborted if user clicks or presses a key during the gesture).

The wall-clock time spent until a lookup was canceled. This is reported every time p2p is used to find a candidate but the request was canceled.

The number of candidates on the LAN, i.e. the number of peers on the LAN offering at least N bytes of the requested file X. This is reported after examining responses from all peers on the LAN and picking a candidate.

The number of p2p downloads of the peer that the returned URL points to. This is reported after examining responses from all peers on the LAN and picking a candidate.

The wall-clock time spent waiting for the LAN-wide number of p2p downloads (i.e. the sum of p2p downloads from each peer on the LAN) to drop below the threshold. This is reported after examining responses from all peers on the LAN and picking a candidate.

The result of the lookup. Possible values include "Found" (if a candidate - i.e. a peer offering at least N bytes of file X - was chosen), "Not Found" (if no candidate could be found), "Vanished" (if a candidate was found but vanished while waiting in line), "Canceled" (if a candidate was found but the request was canceled while waiting in line), and "Filtered" (if it was detected that mDNS was filtered). This is reported after examining responses from all peers on the LAN when p2p is used to find a candidate.

The number of peers implementing p2p file sharing on the network. This is reported every time p2p is used to look up a resource on a network where mDNS is not filtered.

The wall-clock time spent waiting for one or more candidates (i.e. peers offering at least N bytes of file X) that all vanished before the LAN-wide number of p2p downloads dropped below the threshold. This is reported every time candidates were found using p2p but then vanished.

The number of currently connected HTTP clients. This is reported every time a HTTP client connects.

Number of megabytes (1,000,000 bytes) served from the device (via HTTP) where the client disconnects prematurely. This is reported every time a file is served and the client disconnects before receiving all data.

Number of megabytes (1,000,000 bytes) served from the device (via HTTP). This is reported every time a file have been served successfully.

The average speed at which the download was served at, in kB/s. This is reported every time a file have been served successfully.

The number of files available via p2p. This is reported every time a file is added or removed to the /var/cache/p2p directory.

When a client resumes a download, the HTTP request includes range specifier to skip the bytes it already has. This metric conveys this as a percentage of the the file size. This is reported every time a file is served, even if the request does not include a range specifier (in which case 0 is reported).

The result of the HTTP request. Possible values include "Response Sent" (the resource was found and the response was successfully sent), "Response Interrupted" (the resource was found but the client disconnected), "Malformed" (the request was malformed), "Not Found" (the request was for a resource that was not found), and "Index" (the request was for the '/' or '/index.html' resource). This is reported for every HTTP request handled.

The number of times that we try to upload a form that we believe should trigger password generation. False means that something about the form would not allow us to try upload (not an Autofillable field, uploading disabled, Autofill servers in backoff, etc.). True does not mean that the upload actually completed successfully, just that it was started.

The number of accounts stored per site in the the password manager (one event per site)

Deprecated as of Chrome 32. See PasswordManagerActionsTakenWithPsl

Stats documenting how we handle every form containing a password, bucketed by the actions taken.

The total number of sites that the user has blacklisted. Recorded by iterating over stored passwords once per run of Chrome.

Indicates whether the password manager is enabled when a tab is opened. This includes prerendered tabs.

The distribution of responses to the "Do you want Chrome to remember this password"? info bar prompt.

Number of passwords deleted when the user chooses to clear passwords via the clear browsing data UI.

When the user chooses to never remember passwords for a form, we remove all previously saved credentials for that form. This is the count of those credentials.

The password manager only shows those credentials that are considered the best match for a particular form. This stat keep track of the credentials that were not as good of a match and were suppressed.

Indicates whether the user's OS password is blank or not at browser startup.

Breakdown of how other possible usernames are displayed. Recorded every time we autofill a password form.

Breakdown of cases where a password is submitted, but we don't even try and save it. Recorded for every password form submit.

Breakdown on trigger rate of providing a password form autofill entry based on matching stored information using the public suffix list for possible matches.

Indicates whether the save password prompt disappeared in less than one second. This most likely indicates that the prompt was dismissed automatically, e.g. due to a page navigation, before the user was able to respond to the infobar.

Indicates whether the save password prompt was displayed.

Breakdown of which response the user selected from the save password prompt.

The number of times each generated password has been used to log in. Recorded by iterating over stored passwords once per run. This information is persisted and synced.

The number of times each saved password has been used to log in. Does not include generated passwords. Recorded by iterating over stored passwords once per run. This information is persisted and synced.

The number of accounts stored in the password manager (across all sites)

Average CPU utilization of the browser process, read out at each two-minute interval. The utilization is in the 0-100% range per CPU, which is then summed up. I.e. a quadcore system fully loaded would read as 400%.

The number of times a browser process has continuously stayed above a certain threshold of CPU utilization over a certain time period (currently set to two minutes).

Chrome OS (ARM Chromebooks using Exynos 5250 only) Adaptive Support Voltage Group, recorded once per bootup. Indicates which "bin" the SoC is part of, which sets the voltage that different rails on the system will run at. The values 0-11 are valid. A value of 12 indicates an error parsing dmesg and should be investigated. See also Platform.LotIdEnum.

Chrome OS number of disk sectors read at boot from kernel start to login-prompt-ready.

Chrome OS number of disk sectors written at boot from kernel start to login-prompt-ready.

Chrome OS size of allocated swap area in megabytes (before compression)

CPU frequency as percent of the baseline frequency, sampled every 30s. This may be throttled down from 100% due to power dissipation issues (too high temperature). It may also be throttled up (turbo), but the kernel does not report the actual turbo frequency, so we put such samples in the 101% bucket.

Peak total (single core) CPU usage for the last sample interval. The sample interval may vary from seconds to several minutes.

Generic event of interest from Chrome OS. Intended mainly to help assess the frequency of rare error conditions.

Average size of user's Cache directory. Logged once a day, if disk usage is high.

Maximum size of user's Cache directory. Logged once a day, if disk usage is high.

Average size of user's Cache directory. Logged once a day, if disk usage is high.

Maximum size of user's Cache directory. Logged once a day, if disk usage is high.

Average size of user's GCache directory. Logged once a day, if disk usage is high.

Maximum size of user's GCache directory. Logged once a day, if disk usage is high.

Days since the least frequently used account signed in. Logged once a day, if disk usage is high.

Number of users home directories on the device. Logged once a day.

Days since last login of the least recently user on device. Logged once a day, if disk usage is high.

Number of user home dirs on device. Logged once a day, if disk usage is high.

KB in use in the /cache filesystem tree. Logged once a day.

Chrome OS KB in use in the /home/chronos filesystem tree. Logged once a day during log file cleanup.

KB in use in the /data filesystem tree. Logged once a day.

Chrome OS KB in use in the /var filesystem tree. Logged once a day during log file cleanup.

The maximum supported micro-architecture on an Intel platform. This value is logged at program start time.

The 32-bit hash of a kernel warning. This is the hash of the "file:line" string corresponding to the location of the warning, for instance: "/mnt/host/source/src/third_party/kernel/files/drivers /gpu/drm/i915/intel_dp.c:351" (ignore spurious spaces). The hash is produced by this code: while (*string) hash = (hash << 5) + hash + *string++; Separately each warning is also collected (with its hash) via the crash reporter, but only its first occurrence in each boot session. Contact semenzato@ for further info.

Number of logical processors. This includes Hyperthreaded cores.

Chrome OS (ARM Chromebooks using Exynos 5250 only) indication about whether we're part of a special lot ID. Special lot IDs are groups of chips that have special case handling in the kernel for the Adaptive Support Voltage code (the normal logic doesn't work). See also Platform.AsvGroup. Note that fused devices are never part of a special lot (currently) and only some unfused lots are "special".

Chrome OS size of active memory as % of total memory.

Chrome OS active anonymous memory (data segments) as % of total memory.

Chrome OS active file-backed memory (executables, ...) as % of total memory.

Chrome OS size of anonymous memory as % of total memory.

Chrome OS size of buffer cache as % of total memory.

Chrome OS: size of file-backed memory minus swap and buffer cache, as % of total memory.

Chrome OS size of inactive memory as % of total memory.

Chrome OS inactive anonymous memory (data segments) as % of total memory.

Chrome OS inactive file-backed memory as % of total memory.

Chrome OS size of mapped memory as % of total memory.

Chrome OS size of free memory as % of total memory.

Chrome OS size of shared memory in Kbytes.

Chrome OS size of slab memory in Kbytes.

Chrome OS amount of swapped-out memory in Kbytes. These, and all other MEMINFO stats, are snapshotted every 30s.

Chrome OS amount of swapped-out memory as % of total RAM. These, and all other MEMINFO stats, are snapshotted every 30s.

Chrome OS unevictable memory (ramfs, SHM_LOCKED, mlocked) in Kbytes.

Peak memory bandwith (read and write) usage during the last sample interval. The sample interval may vary from seconds to several minutes.

Chrome OS total anonymous memory (active + inactive) as % of total memory 1 minute after boot.

Chrome OS total anonymous memory (active + inactive) as % of total memory 5 minutes after boot.

Chrome OS total anonymous memory (active + inactive) as % of total memory 30 minutes after boot.

Chrome OS total anonymous memory (active + inactive) as % of total memory 150 minutes after boot.

Chrome OS total anonymous memory (active + inactive) as % of total memory 750 minutes after boot.

Page faults per second averaged over 30s interval, sampled continuously.

Page faults per second averaged over 1s interval, sampled every 30s.

Number of disk sectors per second read by Chrome OS in a long interval (currently 30s)

Number of disk sectors per second read by Chrome OS in a short interval (currently 1s, sampled every 30s)

Disk communication errors (SMART 199), sent at boot.

Uncorrectable disk errors (SMART 187), sent at boot.

USB device ID of the charger plugged into a Spring device (if any), sent once a minute. The Device ID is composed from the following 4 8-bit registers of the TSU6721 chip: ADC (07h), Device Type 3 (15h), Device Type 2 (0Bh), Device Type 1 (0Ah). Device Type 1/2/3 is a bitmap and most of bits are mutually exclusive (excepted VBUS debounce). ADC is the 5-bit value of the ID pin, but for most types (as in Device Type), there are only one or two possible ID pin connections/values. The datasheet can be found here: http://www.ti.com/lit/ds/symlink/tsu6721.pdf. Note that different brand/models of the charger can have the same ID.

Chrome OS stateful partition usage level.

Average pages/second swapped IN over a 30s interval, sampled every 30s.

Average pages/second swapped IN over a 1s interval, sampled every 30s.