| Commit message (Collapse) | Author | Age | Files | Lines |
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
* Relax isolation level when indexer reads from DB
This patch causes the indexing subsystem to use snapshot isolation when
reading from the database. This reduces commit conflicts and ensures
the index can make progress even in the case of frequently updated docs.
In the pathological case, a document updated in a fast loop can cause
the indexer to stall out entirely when using serializable reads. Each
successful update of the doc will cause the indexer to fail to commit.
The indexer will retry with a new GRV but the same target DbSeq. In the
meantime, our frequently updated document will have advanced beyond
DbSeq and so the indexer will finish without indexing it in that pass.
This process can be repeated ad infinitum and the document will never
actually show up in a view response.
Snapshot reads are safe for this use case precisely because we do have
the _changes feed, and we can always be assured that a concurrent doc
update will show up again later in the feed.
* Bump erlfdb version
Needed to pull in fix for snapshot range reads.
|
|
|
|
|
|
|
|
| |
Previously, when an erlfdb error occured and a recursive call to `update/3` was
made, the result of that call was always matched against `{Mrst, State}`.
However, in the case when the call had finalized and returned
`couch_eval:release_map_context/1` response, the result would be `ok` which
would blow with a badmatch error against `{Mrst, State}`.
|
| |
|
| |
|
|\
| |
| | |
Optimize lookup/3
|
|/
|
|
|
| |
A tidier version of https://github.com/apache/couchdb/pull/3384 that
saves an unnecessary call to collate.
|
|\
| |
| | |
use collate in lookup
|
| | |
|
| |
| |
| |
| |
| |
| |
| | |
If one of the provided lookup keys doesn't exist in the ebtree, it can
inadvertently prevent a second lookup key from being found if it the
first key greater than the missing lookup key is equal to the second
lookup key.
|
| |
| |
| |
| |
| |
| | |
These two test cases expose the subtle bug in ebtree:lookup_multi/3
where a key that doesn't exist in the tree can prevent a subsequent
lookup key from matching in the same KV node.
|
| | |
|
|\ \
| |/
| | |
Show process status in active_tasks
|
|/
|
|
|
| |
This allows users to verify that compaction processes are suspended
outside of any configured strict_window.
|
| |
|
|
|
|
|
|
|
| |
chunked (#3360)
Transfer-Encoding: chunked causes the server to wait indefinitely, then issue a a 500 error when the client finally hangs up, when PUTing a multipart/related document + attachments.
This commit fixes that issue by adding proper handling for chunked multipart/related requests.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
* Add a development container config for VS Code
This creates a development environment with a FoundationDB server
and a CouchDB layer in two containers, sharing a network through
Docker Compose.
It uses the FDB image published to Docker Hub for the FDB container,
and downloads the FDB client packages from foundationdb.org to provide
the development headers and libraries. www.foundationdb.org is actually
not trusted in Debian Buster by default, so we have to download the
GeoTrust_Global_CA.pem. The following link has more details:
https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=962596
Once the Docker Compose setup is running, VS Code executes the
create_cluster_file.bash script to write down a cluster file containing
the IP address in the compose network where the FDB service can be
found. This cluster file is used both for a user-driven invocation of
`./dev/run`, as well as for unit tests that require a running CouchDB.
Additionally, I've got a small fix to the way we run explicitly specified
eunit tests:
* Run eunit tests for each app separately
The `eunit` target executes a for loop that appears intended to use a
separate invocation of rebar for each Erlang application's unit tests.
When running `make eunit` without any arguments this works correctly,
as the for loop processes the output of `ls src`. But if you specify a
comma-delimited list of applications the for loop will treat that as a
single argument and pass it down to rebar. This asymmetry is
surprising, but also seems to cause some issues with environment
variables not being inherited by the environment used to execute the
tests for the 2..N applications in the list. I didn't bother digging
into the rebar source code to figure out what was happening there.
This patch just parses the incoming comma-delimited list with `sed` to
create a whitespace-delimited list for the loop, so we get the same
behavior regardless of whether we are specifying applications
explicitly or not.
|
| |
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
* Simplify and speedup dev node startup
This patch introduces an escript that generates an Erlang .boot script
to start CouchDB using the in-place .beam files produced by the compile
phase of the build. This allows us to radically simplify the boot
process as Erlang computes the optimal order for loading the necessary
modules.
In addition to the simplification this approach offers a significant
speedup when working inside a container environment. In my test with
the stock .devcontainer it reduces startup time from about 75 seconds
down to under 5 seconds.
* Rename boot_node to monitor_parent
* Add formatting suggestions from python-black
Co-authored-by: Paul J. Davis <paul.joseph.davis@gmail.com>
|
| |
|
| |
|
|
|
| |
1. The caching effort was a bust and has been removed. 2) chunkify can be done externally with a custom persist_fun.
|
|
|
|
|
| |
All endpoints but _session support gzip encoding and there's no practical reason for that.
This commit enables gzip decoding on compressed requests to _session.
|
| |
|
| |
|
| |
|
| |
|
| |
|
|\
| |
| | |
Add ability to control which Elixir integration tests to run
|
| | |
|
|/
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
New `elixir-suite` Makefile target is added. It runs a predefined set of elixir
integration tests.
The feature is controlled by two files:
- test/elixir/test/config/suite.elixir - contains list of all available tests
- test/elixir/test/config/skip.elixir - contains list of tests to skip
In order to update the `test/elixir/test/config/suite.elixir` when new tests
are added. The one would need to run the following command:
```
MIX_ENV=integration mix suite > test/elixir/test/config/suite.elixir
```
|
|
|
|
| |
Add missing default headers to responses
|
| |
|
|
|
|
|
|
|
|
|
|
| |
This flips the view indexer to grab the database update_seq outside of
the update transaction. Previously we would cosntantly refresh the
db_seq value on every retry of the transactional loop.
We use a snapshot to get the update_seq so that we don't trigger
spurious read conflicts with any clients that might be updating the
database.
|
|
|
|
|
|
|
| |
This is useful so that read conflicts on the changes feed will
eventually be resolved. Without an end key specified a reader could end
up in an infinite conflict retry loop if there are clients updating
documents in the database.
|
| |
|
|\
| |
| | |
Fix mango tests
|
| | |
|
|/ |
|
|
|
|
| |
https://github.com/apache/couchdb-erlfdb/releases/tag/v1.2.3
|
|\
| |
| | |
Use `req_body` field if present
|
|/
|
|
|
|
|
| |
When we call `couch_httpd:json_body/1` we can have `req_body` already set.
In this case we should return the field as is without any attempt to
decompress or decode it. This PR brings the approach we use in `chttpd`
into `couch_httpd`.
|
|\
| |
| | |
Avoid deleting UUID keys that start with zeros
|
| | |
|
|/
|
|
|
| |
Any ebtree that uses chunked key encoding will accidentally wipe out any
nodes that have a UUID with more than one leading zero byte.
|
|
|
|
|
|
|
|
|
| |
Waiting for the timeout option to be set means we could still sneak in
and grab the old FDB database handle before fabric2_server updated it in
the application environment.
This new approach just waits until the handle has been updated by
watching the value in the application environment directly.
|
|
|
|
|
| |
Turns out that ebtree caching wasn't quite correct so removing it for
now.
|
|
|
|
|
|
|
|
| |
The ebtree caching layer does not work correctly in conjunction with
FoundationDB transaction retry semantics. If we incorrectly cache nodes
that are not actually read from FoundationDB, a retried transaction will
rely on incorrectly cached state and corrupt the ebtree persisted in
FoundationDB.
|
|\
| |
| | |
Add an "encryption" object to db info
|