summaryrefslogtreecommitdiff
Commit message (Collapse)AuthorAgeFilesLines
* WT-2403 Remove custom_die from new test.Michael Cahill2016-10-141-1/+0
|
* WT-2403 Add next_random to LSM cursors (#2887)David Hows2016-10-1410-88/+529
|
* WT-2955 Add statistics tracking the amount of time threads spend waiting for ↵Keith Bostic2016-10-1315-258/+531
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | high level locks (#3086) * WT-2955 Add statistics tracking the amount of time threads spend waiting for high level locks Sort the statistics categories so it's easier to find stuff, no real change. * Add counters and usec wait times to long-term locks (currently the checkpoint, handle-list, metadata, schema and table locks). * mutex.i:295:26: error: conversion to int64_t {aka long int} from long unsigned int may change the sign of the result [-Werror=sign-conversion] [t->slot_usecs] += WT_TIMEDIFF_US(leave, enter); * Rename the lock statistics so they group together. Split lock wait times between internal and application threads. * Separate the connection's dummy session initialization out into its own function, that way it's clear what we're doing. * The session's connection statistics are fixed when the session ID is allocated, so we can cache a pointer to them and avoid u_int divisions (which are currently about the slowest thing you can do on a modern architecture). * A slightly different change: instead of caching a reference to the connection statistics, cache the offset into the array of statistics pointers, that way we can avoid the integer division when updating almost all statistics. * Review comments: Add comments describing the use of statistics array offsets in lock tracking. Rename WT_STATS_FIELD_TO_SLOT to WT_STATS_FIELD_TO_OFFSET. Whitespace cleanup. * __wt_cache_create() doesn't need to call __wt_cache_destroy() explicitly, if the connection open fails at any point, __wt_cache_destroy() will be called as part of that cleanup. * Append the suffix "_off" to the spinlock structure statistics field names, clarify they're offsets into the statistics structures.
* WT-2965 Remove sleep and retry loop from __evict_exclusive (#3091)David Hows2016-10-121-11/+2
|
* WT-2949 Add option to skip closing the connection after test runs. (#3082)sueloverso2016-10-117-1/+14
| | | | | | | | | * WT-2949 Add option to skip closing the connection after test runs. * Reverse the default value per review comments and add where needed to configs. * Add additional text/comments for new option.
* WT-2963 Fix a race bounding eviction intiny caches. (#3090)Michael Cahill2016-10-111-7/+4
|
* WT-2959 Ensure WT_SESSION_IMPL is never used before it's initialized (#3087)Keith Bostic2016-10-113-13/+13
| | | This is not a bug fix, it's a change to make the code safer against future misuse.
* WT-2956 utility tests -h option is always overridden by the default setup ↵Keith Bostic2016-10-1030-78/+18
| | | | (#3085)
* WT-2954 Evict pages that exceed memory_page_max even if the transaction ID ↵Alex Gorrod2016-10-102-12/+39
| | | | | | | | | | | is stuck (#3088) Try to evict pages that exceed memory_page_max even if the transaction ID isn't increasing. Workloads that insert large values checking only for the transaction ID moving forward can lead to very large in-memory pages. Allow the page to exceed the split size, but not the configured maximum. Skipping such pages is a performance optimization - allow the optimization to work as long as pages are between split size and maximum size. Also don't require so many entries on a leaf page to trigger a split when there are large values.
* WT-2937 Fix a warning. (#3084)Michael Cahill2016-10-071-4/+4
| | | | | | | | | | | ../src/evict/evict_lru.c:1081:29: error: implicit conversion loses integer precision: 'unsigned long long' to 'u_int' (aka 'unsigned int') [-Werror,-Wshorten-64-to-32] WT_MIN(max_entries, 1 + __wt_cache_pages_inuse(cache) / 2); ~~^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ ../src/include/misc.h:62:42: note: expanded from macro 'WT_MIN' ^ ../src/evict/evict_lru.c:1084:29: error: implicit conversion loses integer precision: 'unsigned long long' to 'u_int' (aka 'unsigned int') [-Werror,-Wshorten-64-to-32] WT_MIN(max_entries, 1 + cache->pages_dirty_leaf / 2); ~~^~~~~~~~~~~~~~~~~~~~~~~~~~~~~ ../src/include/misc.h:62:42: note: expanded from macro 'WT_MIN'
* WT-2950 Free all reconciliation memory between calls to eviction. (#3081)Michael Cahill2016-10-071-11/+10
|
* WT-2937 Only do an eviction walk after a page has been taken from the ↵Michael Cahill2016-10-071-36/+38
| | | | | | | | | | | | | | | current queue. (#3071) Along with a number of other changes to avoid pathological cases with tiny caches: Application threads should do urgent eviction if there is only one eviction worker. Always alternate the "fill queue", so that once a walk has tried to fill one queue, it keeps evicting from the other. If there are tiny numbers of candidate pages, only allow half in each queue to avoid situations where all candidates are in one queue. If there is only one candidate page in the whole cache, it had better get queued. Don't allow the eviction server to sit on pages requiring urgent eviction.
* WT-2932 Add a configuration option allowing tables to ignore cache limits ↵Alex Gorrod2016-10-077-36/+140
| | | | | | | (#3069) Only available for in-memory configurations, designed to be used for applications that maintain a small amount of content that is required for correctness.
* WT-2948 simplify error handling by making __wt_epoch return never fail (#3080)Keith Bostic2016-10-0734-218/+145
| | | If a system call to retrieve a timestamp fails it will now result in a panic. We couldn't find any case where that's a real possibility.
* WT-2831 Skip checkpointing if there have been no modifications (#3067)Sulabh Mahajan2016-10-0611-35/+94
| | | This saves locking and scanning the list of active handles for read-only workloads, which can be time consuming when there are many tables.
* WT-2946 dist/s_docs incompatible with OS X Xcode installation (#3079)Keith Bostic2016-10-041-7/+3
| | | Gcc/clang both take - to mean stdin, get rid of the temporary file.
* WT-2920 New eviction statistics (#3076)Keith Bostic2016-10-045-32/+70
| | | | | | | | | * WT-2920 New eviction statistics Add "application_eviction" statistic, the usecs application threads spent doing eviction. * Add a finer grained timer around eviction as well.
* WT-2946 dist/s_docs incompatible with OS X Xcode installation (#3078)Keith Bostic2016-10-041-4/+11
| | | | | | | | The OS X cpp program injects line number output in the middle of lines and doesn't stringify #XXX entries; use the -E option to the compiler instead. The OS X C pre-processor won't take input from stdin, use a temporary source file.
* WT-2938 Change assembly file extensions from .S to .sx (#3072)David Hows2016-10-045-4/+4
|
* WT-2941 Improve test/format to use faster key-generation functions (#3073)Keith Bostic2016-10-041-6/+16
|
* WT-2933 Fix an uninitialized variable warning. (#3077)Michael Cahill2016-10-041-2/+4
|
* WT-2933 Fix a race between named snapshots and checkpoints. (#3068)Michael Cahill2016-10-0411-57/+110
|
* WT-2942 Verbose messages should not have newlines. (#3074)sueloverso2016-09-306-15/+14
| | | | | | | * WT-2942 Verbose messages should not have newlines. * Fix another set of ERR/RET macros and __wt_verbose calls including newline characters.
* WT-2907 Fixed java concurrent close test to have both insert threads and ↵Don Anderson2016-09-281-10/+11
| | | | scan threads. (#3066)
* WT-2895 Reduce the runtime of make check testing with disable long (#3052)David Hows2016-09-285-0/+11
| | | Mostly in order to allow valgrind runs with make check to finish in a reasonable time.
* WT-2858 Rename wtperf's CONFIG structure (#3065)Keith Bostic2016-09-2811-745/+814
| | | | Also replace key-string generation implementation with a more efficient implementation.
* WT-2931 Enforce limits on dirty data for in-memory. (#3064)Michael Cahill2016-09-273-47/+39
| | | Reverts commit ac1f7401dcb8be345973f7787d9121c5c321bf7b.
* WT-2917 split wtperf's configuration into per-database and per-run parts (#3006)Keith Bostic2016-09-2716-869/+826
|
* WT-2928 Don't give up when the urgent queue is half empty. (#3063)Michael Cahill2016-09-261-4/+4
|
* WT-2928 Eviction failing to switch queues can lead to starvation (#3060)Keith Bostic2016-09-261-40/+58
| | | There are some configurations where the eviction server waiting for a the current queue to drain before switching in the new queue can lead to starvation. Specifically if in-memory is configured, the cache is full and there is only a single eviction thread. Or if an operation that does not contribute to eviction fills the cache (for example compact).
* WT-2921 Reset the count when the last hazard pointer is cleared. (#3058)Michael Cahill2016-09-261-2/+3
|
* WT-2924 Ensure we are doing eviction when threads are waiting for it (#3061)Keith Bostic2016-09-253-17/+18
| | | | | | * Set WT_CACHE_EVICT_XXX when setting WT_CACHE_EVICT_XXX_HARD (the byte count checks can race, so we have to explicitly set both). * Make WT_CACHE.pages_evict volatile: loops in __evict_pass() and __wt_cache_eviction_worker() read the value and depend on it being updated when subsequently read.
* WT-2924 Ensure we are doing eviction when threads are waiting for it. (#3056)Michael Cahill2016-09-234-35/+77
| | | Fix a bug in scrub step down, where it could set the target below the eviction target leading to an unexpected eviction state.
* WT-2919 Don't mask error returns from style checking scripts (#3049)Keith Bostic2016-09-2320-19/+41
| | | We were forcing them to return an exit code of 0.
* WT-2926 WT_CONNECTION.reconfigure can attempt unlock of not-locked lock (#3057)Keith Bostic2016-09-231-2/+6
| | | Don't unlock the spin lock unless we've locked it.
* WT-2925 WT_THREAD_PANIC_FAIL is a WT_THREAD structure flag (#3055)Keith Bostic2016-09-231-7/+8
| | | | | WT_THREAD_PANIC_FAIL is a WT_THREAD structure flag, but is declared elsewhere, likely to introduce a bug if additional flags are added to the WT_THREAD structure.
* WT-2864 Update reconfigure test to detect hangs (#3051)Keith Bostic2016-09-231-11/+42
| | | | | | | Set an alarm when running reconfiguration tests so we can detect hangs; if the alarm goes off, output the current configuration string and drop core. We've seen cases where reconfiguring causes a hang, but haven't been able to identify the root cause.
* WT-2923 heap-use-after-free on address in compaction (#3053)Keith Bostic2016-09-212-30/+51
| | | We can't look at the WT_REF address without some kind of lock as the address can change underneath us. For example, if we take a copy of WT_REF.addr.addr while it's referencing an on-page cell, and then the page is evicted before we crack that address, we'll potentially read freed memory. Lock down the WT_REF before reading the address.
* WT-2874 Change test_compact01 to avoid eviction (#3050)Keith Bostic2016-09-191-3/+5
| | | The test relies on operating without eviction.
* WT-2918 the dist scripts create C files s_whitespace complains about (#3048)Keith Bostic2016-09-163-25/+30
| | | | | | | | | | | | | | | | | | | | | | | | | | | * WT-2918 the dist scripts create C files s_whitespace complains about s_export doesn't need to create a .c file. enhance a comment to explain the loop. * The s_tags script creates a .c file that s_whitespace doesn't like, use a standard source file for the ctags flag test. At the end of the script, remove the tags file created in the dist directory. Include source files deeper than 1 directory below src/ so we get src/checksum/XXX. Don't redirect the "skipping ctags" message to the tags file, that means it never gets seen. * Fixing the bug that kept the s_tags "skipped" message from showing up broke s_all, add s_tags to the exclusion list. * Don't overwrite the local tags file if we're run in the wrong directory. Exit 0 on success, 1 on error.
* WT-2916 Fix and simplify s_whitespace. (#3047)Michael Cahill2016-09-1696-170/+46
|
* WT-2910 When running in-memory, only evict dirty pages. (#3044)Michael Cahill2016-09-162-5/+42
| | | | | * Don't relax the clean / dirty requirements when we get aggressive: if application threads are blocked by the dirty limit, evicting a clean page doesn't unblock them so there's no point trying. * Sanity check dirty eviction limits: once the cache is full, we should evict both clean and dirty pages.
* WT-2905 dead code: MongoDB Coverity #99881. (#3045)Michael Cahill2016-09-161-2/+2
|
* spellingMichael Cahill2016-09-151-4/+4
|
* WT-2913 Re-sort eviction queues if we find no new candidates. (#3043)Michael Cahill2016-09-151-22/+20
| | | | | | We could be in a situation with a tiny number of pages in cache, all of which are already queued. The queue needs to be re-sorted so that some of those entries can become candidates. This caused hangs in some test/format configs.
* WT-2914 In test/csuite program, added explicit call to wiredtiger_open to ↵Don Anderson2016-09-141-5/+11
| | | | satisfy some linkers (#3040)
* WT-2859 Restructure statistics gathering macros (#3016)Sulabh Mahajan2016-09-1456-455/+433
| | | | | | | Having _FAST_ macros gives an impression that when we use them, we are collecting fast statistics only, which is not true. Except when statistics=none is set, we collect all the stats. This change removes _FAST_ macros and modifies the basic macros to only collect stats when statistics=none is not set.
* WT-2902 Track per-tree the efficiency of eviction walks. (#3036)Michael Cahill2016-09-1410-221/+271
| | | | | | | | | Some workloads create "deserts" within a tree where no good eviction candidates can be found (e.g., zipfian distributions, append-only workloads, etc.). Rather than walking all the way through such regions, give up and try in the opposite direction next time. Also add a statistic for the eviction empty score.
* WT-2911 add support for gcc6 (#3037)Keith Bostic2016-09-133-11/+23
| | | | | | | | | | | | | * WT-2911 add support for gcc6 * Turn on stronger clang flags except for Apple's clang 4.1. Another run at getting the quoting right. * rebalance.c:81:2: error: embedding a directive within macro arguments has undefined behavior [-Werror,-Wembedded-directive] * Remove double quotes around string in case statements, it makes bash sad.
* WT-2912 make --enable-strict run on the zseries test box (#3038)Keith Bostic2016-09-141-0/+4
| | | | | | | | gcc 4.8.5 on the zseries (incorrectly) thinks callers of __wt_block_write_off() can use uninitialized size, offset and checksum returns, but those variables are only left uninitialized on error return. gcc 4.8.5 (Red Hat 4.8.5-4) Linux rhel72-z-wt.novalocal 3.10.0-327.4.4.el7.s390x #1 SMP Thu Dec 17 15:52:10 EST 2015 s390x s390x s390x GNU/Linux
View Lorry Controller Status