| Commit message (Collapse) | Author | Age | Files | Lines |
|\
| |
| |
| |
| |
| |
| |
| |
| |
| | |
* stable-3.2:
Make the indexing operation fail upon StorageException(s)
ConsistencyCheckerIT: Delete calls to index broken changes that fail silently
RefAdvertisementIT: Don't call reindex that would fail silently
Release-Notes: skip
Change-Id: Iecdf1e4ce450c8fab5244338ce46e6e4d95bf959
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Change 270450 caused the blanking of the Lucene document
upon reindexing if any field caused a StorageException.
Whilst the overall intention was good, the implementation
caused the Lucene index replace operation to continue with
a Document without any fields instead of making the whole
operation fail.
StorageExceptions are thrown when the underlying storage,
being a filesystem or anything else, returns some errors.
Whether the failure is permanent or temporary (e.g. concurrent
GCs, repacking or pruning may cause sporadic StorageException(s))
returning a blank Lucene document was incorrect because, instead
of failing the operation, it was causing the change entry to be
completely removed from the index.
Let the StorageException fail the indexing operation, so that
existing entries may continue to exist, allowing the caller
to retry the operation.
The previous implementation, returning an empty Document, did
not allow any retry, because once the change entry was removed
from the index, it could not be discovered and reindexed anymore
for example by a `gerrit index changes <changenum>`.
Tested manually, applying a randomic StorageException thrown
during the fetching of the ChangeField extensions:
public static Set<String> getExtensions(ChangeData cd) {
if (new Random().nextBoolean()) {
throw new StorageException("Simulated storage exception");
}
return extensions(cd).collect(toSet());
}
Before this change, every time one change indexing throws a
StorageException, it disappears from the index. Eventually,
all changes will be eliminated and only an off-line reindex
or the reindex of all changes in the project would allow to
recover them.
After this change, some of the indexing operations are
successful and other fails. However, retrying the reindexing
operation would allow to reindex all of them.
Even if the above test case looks strange at first sight,
it simulates a real-life scenario where a low percentage
of indexing operation (0.1%) may fail because of sporadic
StorageExceptions. Before this change, some index entries
were missing on a daily basis (5 to 10 changes per day),
whilst after this change all indexing operation can be
retried and do not result in any indexing entry loss.
Bug: Issue 314113030
Release-Notes: Fail the change reindex operation upon StorageException(s)
Change-Id: Ia121f47f7a68c290849a22dea657804743a26b0d
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
These calls used to fail silently and are of no further use.
This change is part of a series that allows for the fake change index
to run in integration tests. The fake index throws an exception in
case reindexing fails, which made this test fail.
Release-Notes: skip
Change-Id: I9c9d3f7613bca057bf0963442077bb6c0b3dab51
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
receivePackOmitsMissingObject wants to test a scenario where we
have a change ref, but the patch set ref / commit was deleted or
is otherwise gone.
It did that by adding a new patch set pointing to a non-existent
SHA1 and triggered a reindex of that change. However, that reindex
fails silently because we submit the reindex task to another
executor and provide no feedback to the caller.
This commit modifies the test and just deletes the patch set ref
which makes for the same scenario but without the reindex trouble.
This change is part of a series that allows for the fake change index
to run in integration tests. The fake index throws an exception in
case reindexing fails, which made this test fail.
Release-Notes: skip
Change-Id: Icf79795a343a6abca28c6504b279eb0a5c84fad2
|
|\|
| |
| |
| |
| |
| |
| |
| | |
* stable-3.2:
Update Jetty to 9.4.53.v20231009 for security updates
Release-Notes: skip
Change-Id: If1d4dfc04d9407f0bab937a996d9321358a7b6a8
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Jetty 9.4.53.v20231009 includes the following two security fixes:
- CVE-2023-36478 [1] - zero-days security issue discovered on
the 10th of October, also known as "HTTP/2 Rapid Reset"
- CVE-2023-44487 [2] - HTTP/2 Stream Cancellation Attack
[1] https://nvd.nist.gov/vuln/detail/CVE-2023-36478
[2] https://nvd.nist.gov/vuln/detail/CVE-2023-44487
Release-Notes: Update Jetty to 9.4.53.v20231009 with critical security fixes
Change-Id: Ie93fbcb8b35d9e4997dc0578893a8856b56b173c
(cherry picked from commit c49057e05d35ff2ad1a7307aa9168b84ae7588db)
|
|\|
| |
| |
| |
| |
| |
| |
| |
| | |
* stable-3.2:
Introduce cache.threads option to enable a custom cache executor
GitwebServlet: Fix project root computation
Release-Notes: skip
Change-Id: I342a210680ff43796f9fa4403f1b70f78e0ddb3e
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Since the introduction of Caffeine as alternative to Guava
in Change 244612, the execution of the cache event listeners moved
to a background thread, run by the ForkJoinPool's common pool [1].
The subtle difference has caused issues to the plugins that are
registering listeners, like the high-availability and multi-site:
the consequences have been quite serious because of the inability
to understand if the eviction was caused by a forwarded cache
eviction event or by an execution of Gerrit API or other internals
that caused the removal of the entry.
The use of the JVM common pool has several disadvantages and, under
certain conditions [2], it may even lead to deadlocks or unexpected
blockages.
By introducing the cache.threads option, decouple the cache background
threads execution and allow to configure an explicit separate
thread pool which can be tuned and decoupled from the rest of the
JVM common threads.
Also, allow to restore the plugins' cache listeners legacy behaviour
without losing the ability to leverage the performance of Caffeine
cache vs. the traditional Guava.
By default, this change is a NO-OP because it preserves the current
behaviour of background execution tasks of the Caffeine cache.
Introduce DefaultMemoryCacheFactoryTest class from stable-3.4 for
avoiding further conflicts when merging upstream.
References:
[1] https://github.com/ben-manes/caffeine/wiki/Guava#asynchronous-notifications
[2] https://dzone.com/articles/be-aware-of-forkjoinpoolcommonpool
Release-Notes: introduce cache.threads option to allow custom executors for Caffeine caches
Bug: Issue 16565
Change-Id: I204abd1bdbf2bbed5b3d982d14cbc5549ac96ace
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
If9da362140 introduced regression in project root computation.
Also add test coverage for Gitweb servlet so that a similar regresssion
could be avoided in the future.
Test Plan:
bazel test javatests/com/google/gerrit/httpd/...
Bug: Issue 16449
Release-Notes: Fix project root computation in Gitweb servlet
Change-Id: Ia1a7bdea55db4deb55a3b82a292890e7febbe675
|
|\|
| |
| |
| |
| |
| |
| |
| | |
* stable-3.2:
Disable printing cache stats by default on Init/Reindex
Release-Notes: skip
Change-Id: Ia8356e393bd8cd8601e4805a0a7106b7a047b593
|
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Printing cache statistics can be slow, particularly when the
caches are large. For example, when the caches are ~400G,
printing cache stats takes over 30 mins.
Release-Notes: Printing cache stats is disabled by default for Init and Reindex
Change-Id: I848d6dfef0be9dee9ebac8a597dd7e617cbe30e4
|
|\|
| |
| |
| |
| |
| |
| |
| |
| | |
* stable-3.2:
Predicate: Add safety check against not(any())
Add an index-safe predicate that matches all changes
Release-Notes: skip
Change-Id: Ie1006ba75c63ef445771acea26dcd371eb750156
|
| |\
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
* stable-3.1:
Predicate: Add safety check against not(any())
Add an index-safe predicate that matches all changes
Release-Notes: skip
Change-Id: I56708f2d9a3218f2981e65b867003d0ea714407a
|
| | |\
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
* stable-3.0:
Predicate: Add safety check against not(any())
Add an index-safe predicate that matches all changes
Release-Notes: skip
Change-Id: Id37339b406675721a4bac3411c6e2874986954bb
|
| | | |\
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | | |
* stable-2.16:
Predicate: Add safety check against not(any())
Add an index-safe predicate that matches all changes
Release-Notes: skip
Change-Id: Ia64e074194cd561bc55a6f0e7976633aaa9bb2ff
|
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | | |
It's tempting to use as a way to say "match no results", but it's very
unsafe. This simple check won't catch all misuses, but it would have
prevented Iad564dd47 from causing production problems.
Fix one obvious caller that now trips this check. This was a real
problem: a bare [is:watched] query on a large site for a user with no
project watches required iterating all changes to return an empty
result. We have real instances of this causing extremely long requests
in our server logs.
Release-Notes: skip
Change-Id: Ib43f7466db5db315bf87558a37f18a1832ec641f
(cherry picked from commit 02bab2161cd7ce7f682a28dad496a4465cb40844)
|
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | | |
Iad564dd47 was problematic in the error case because it returned
Predicate.not(Predicate.any()). The "any" predicate is not an index
predicate, which produces at least two problems:
* When used in a with another non-index predicate, it
might throw "no ChangeDataSource" from AndSource/OrSource.
* More dangerously, when combined with an index predicate, it is used as
a post-filtering predicate. A post-filtering predicate that matches no
results means it gets applied to every change in the index before
returning an empty set, which is bad.
Instead, reuse a handy existing index predicate that is designed to
return no results efficiently from the index:
ChangeStatusPredicate.NONE. Expose this from a more general location,
ChangeIndexPredicate#none().
Release-Notes: skip
Change-Id: Ic9a7e277070cff7768ec91c1972cf8e9f6deacb1
(cherry picked from commit d9679cdb605b8a10977e1fc821f57dd5749b1dd5)
|
|\| | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | | |
* stable-3.2:
Cache repository locations in LocalDiskRepositoryManager
Release-Notes: skip
Change-Id: I4052ceea68399279f23a0e180227b71643e9a579
|
| |\| | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | | |
* stable-3.1:
Cache repository locations in LocalDiskRepositoryManager
Release-Notes: skip
Change-Id: I5443e5174552f2a8369f5d97162b9cb24cd2ae26
|
| | |\| |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | | |
* stable-3.0:
Cache repository locations in LocalDiskRepositoryManager
Release-Notes: skip
Change-Id: I5b6cd9e19c4a0b052da9705e0a8ddb0eced148c2
|
| | | |\|
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | | |
* stable-2.16:
Cache repository locations in LocalDiskRepositoryManager
Release-Notes: skip
Change-Id: I39713079d081e7b80f59dc36e6181f4daa565c50
|
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | | |
Obtaining the actual location of a repository using base-path
and project name can be slow as it involves some guessing to
locate the repository. Cache the locations once they are
obtained to avoid repeated work, thereby improving performance
when opening repositories.
For example, on a site with 20k repositories on NFS, ls-projects
with this change takes ~60s and ~90s without. Also, a query which
wraps a large (~2k) list of manifest[1] operators will take ~300ms
with this change and ~2s without it.
[1] https://gerrit.googlesource.com/plugins/manifest
Release-Notes: skip
Change-Id: I8eab3c813c4ac9433e93c7ace96d38efe332be27
|
|\| | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | | |
* stable-3.2:
Limit the number of changes that can be submitted together
GitwebServlet: Retrieve git path from FileRepository
Make delegate() method public
Release-Notes: skip
Change-Id: I56968d4a2cf9800e78e673042a1cc250879b3fd8
|
| |\| | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | | |
* stable-3.1:
Limit the number of changes that can be submitted together
Release-Notes: skip
Change-Id: Iab4d9bbbb44cb4e5e3db084f4350e1079f0e7195
|
| | |\| |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | | |
* stable-3.0:
Limit the number of changes that can be submitted together
Release-Notes: skip
Change-Id: I63765e81c80fa3bfb661c11ada5b7013f93f93e1
|
| | | |\|
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | | |
* stable-2.16:
Limit the number of changes that can be submitted together
Release-Notes: skip
Change-Id: I263d636ec38f043ad5f6f8157ea5a57e12e7b145
|
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | | |
When chaining changes together, the sequence of commits to navigate
was previously unbound, causing the potential explosion to millions
of changes.
The explosion could have also been accidental and caused by the push
of a change with a non-existent branch, which would have resulted
in the full scan of the repository for changes.
Introduce a new Gerrit configuration change.maxSubmittableAtOnce with
a safe default of 32767, which would allow any use case that would have
also worked before this change. Navigating over 32767 changes up to
potentially a huge number of commits would have generated a significant
CPU and memory overload and still not resulted in a submittable
chain of changes anyway.
Release-Notes: Limit the number of changes that can be submitted at once
Bug: Issue 16322
Change-Id: Id71aed2341f72708778395359bb6e4d4c270401c
|
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | | |
Instead of retrieving git base path from repository manager, open the
repository and get it directly from its git directory path.
Also, redirect the delegate repository wrapper to the underlying
implementation, allowing to use GitWeb in combination with the
multi-site plugin and the cached-refdb.
The advantage of doing it: we don't need to cast GitRepositoryManager to
the LocalDiskRepositoryManager, so that other implementations would also
work.
Release-Notes: allow using GitWeb with multi-site and cached-refdb
Change-Id: If9da36214063d73953677473082cd16f8f95163a
|
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | | |
The delegate() method was initially introduced in I8a862ac852
to allow access to the underlying repository while running GC
for sites using multi-site plugin.
The method was initially package private since the callers
where in the same package.
Changing visibility of delegate() method to public
to allow plugins to access it.
Accessing the wrapped repository would otherwise require hacky
workarounds like this Ia3fbdc7e96.
Bug: Issue 15997
Release-Notes: Make DelegateRepository#delegate() method public
Change-Id: I4c8b4f97f9995a7adeca5b6f763e7913c9e8a5e5
|
|\ \ \ \ \ |
|
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | | |
this update to current stable-5.9 contains the following changes:
fed0ab9ba Use FileSnapshot without using configs for FileBasedConfig
7828ef349 Revert "RefDirectory.scanRef: Re-use file existence check done in snapshot creation"
994434a1c Add missing @since tags
2e5110eda Add missing @since tag
8924b14d3 Add missing @since tags
f8f4357d6 Remove unused import in ApacheSshTest
f1547eec0 Update maven plugins
fe3071f0e Ignore missing javadoc in test bundles
38db89142 storage: file: De-duplicate File.exists()+File.isFile()
f829f5f83 RefDirectory.scanRef: Re-use file existence check done in snapshot creation
5606a5315 FileSnapshot: Lazy load file store attributes cache
09c923073 Update eclipse-jarsigner-plugin to 1.3.2
c213a6628 Fix p2 repository URLs
122237439 FS: debug logging only if system config file cannot be found
6aa29d116 Better git system config finding
4f8d43462 Fix target platforms
d160e8a93 Fix missing peel-part in lsRefsV2 for loose annotated tags
5f8c48413 reftable: drop code for truncated reads
b4782d74f reftable: pass on invalid object ID in conversion
9c3190ce7 Update eclipse-jarsigner-plugin to 1.3.2
283c23012 Fix running benchmarks from bazel
c70c0acb4 Update eclipse-jarsigner-plugin to 1.3.2
d160e8a93 should fix Issue 14861
Bug: Issue 14861
Release-Notes: Update jgit to c6b0ee04e49c96e0beec4154196c416abcf2bcc9
Change-Id: Ia221acf855dfc95816def73d77af212888d700a9
|
|\ \ \ \ \ \
| | |/ / / /
| |/| | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | | |
* stable-3.2:
Detect DelegateRepository in GarbageCollection operation
Allow reuse of DelegateRepository functionality
Release-Notes: skip
Change-Id: I1da2532d1d45b7bf3af0dddea821168eb897cd04
|
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | | |
When GarbageCollectCommand performs GC it expects that instance of the
repository that is passed to it is either FileRepository or
DfsRepository and throws UnsupportedOperationException otherwise.
Detect if DelegateRepository is passed as parameter and make sure that
check is performed on a delegated repository instead.
Bug: Issue 14945
Release-Notes: skip
Change-Id: I8a862ac852f5f98c09662a41234d40c26e944804
(cherry picked from commit a02af610c8ec879c79259a5d4b7a185a22311cd1)
|
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | | |
By allowing plugins to reuse DelegateRepository functionality
multi-site and high-availability plugins can avoid code
duplication and significantly reduce the code complexity.
Also, make the annotation @UsedAt repeatable for associating
the DelegateRepository functionality with the usage from multiple
plugins.
Bug: Issue 13429
Release-Notes: skip
Change-Id: I2b5f5b215395fc1ef2a8008a71f5c09278d1278b
(cherry picked from commit ed54a267dafeed829722b340ccaee2bbbc7f94b3)
|
|\| | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | | |
* stable-3.2:
Allow async receive-commits to have a thread-local cache
Fix RepoRefCache stale checks during NoteDb rebuild
Lazy load change notes when submit by push
Change-Id: Idd341667dcc1d6ba9ffe7423776d55b3de659eb5
Release-Notes: skip
|
| |\ \ \ \ \
| | | |/ / /
| | |/| | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | | |
* stable-3.1:
Allow async receive-commits to have a thread-local cache
Fix RepoRefCache stale checks during NoteDb rebuild
Lazy load change notes when submit by push
Change-Id: Idb781e53343f98c898ecd4cd1033160b4ce4dfae
Release-Notes: skip
|
| | |\ \ \ \
| | | | |/ /
| | | |/| |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | | |
* stable-3.0:
Allow async receive-commits to have a thread-local cache
Fix RepoRefCache stale checks during NoteDb rebuild
Lazy load change notes when submit by push
Change-Id: Ie82487167fc84ec543083b018c3fde1b0a041e1d
Release-Notes: skip
|
| | | |\ \ \
| | | | | |/
| | | | |/|
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | | |
* stable-2.16:
Allow async receive-commits to have a thread-local cache
Fix RepoRefCache stale checks during NoteDb rebuild
Lazy load change notes when submit by push
Change-Id: I52fd7c9320ce57488ab54f8b5b24a8eabb002ea5
Forward-Compatible: checked
Release-Notes: skip
|
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | | |
Git receive-commits are executed in a background thread of the
ReceiveCommits pool. The thread-local cache allocated on the
client caller thread isn't used in the execution of the command
making multiple operations (e.g. ACL evaluation, events propagation)
slower because of the missed caching activity.
Release-Notes: Improve caching when merging changes through git push
Change-Id: I34a6e1485294f3156c7f35261fedfc280af1ed43
|
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | | |
When rebuilding NoteDb meta-data from ReviewDb we are
actually mutating the /meta ref during a reindexing
operation.
This is an edge case that happens only once during
the on-line NoteDb upgrade and the trial mode.
Adapt the RepoRefCache use during NoteDb reindexing
and manage the situation where a staleness check
applies to a ref that has disappeared on disk.
Bug: Issue 15961
Release-Notes: skip
Change-Id: Iac642eb7a9b24882a0c77dbaee24853fb0b29827
|
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | | |
Lazy load the change notes associated with the change to merge
when one or more changes are submitted by a git push of
the change onto its target branch.
When a single change is merged through git push, we need to find
which one is the corresponding change meta to be checked and
merged. However, the building of the entire set of open changes
against the target branch, indexed by their change key was having
the side-effect of loading eagerly changes notes from their /meta ref.
Loading all changes notes can be very slow, O(mins), on
repositories with a large number of refs on slow storage and, also,
it is completely unneeded in this use case.
Loading all the change notes for all open changes had increased the
time to submit a change from a few seconds (via the GUI) to a
few minutes (via git push).
Release-Notes: Performance improvement of change submit via push
Change-Id: I646bc7b13f3359816b6be6354e15ddb5412bfa38
|
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | | |
In SSH scope, we are creating an IdentifiedUser each time we try to
get the user. In I141e9bf2, it seems that the intention was to create
the identified user only the first time it was asked for. Creating a
new IdentifiedUser each time adds an overhead that degrades query
performance.
We do want the IdentifiedUser created via the RequestFactory so that
request scoped fields are populated correctly which is why we don't
directly return CurrentUser#asIdentifiedUser().
Release-Notes: SSH query performance is improved
Change-Id: Ibca6c2320ed3abd7bfa6bf29daaacb5048d6a798
|
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | | |
Use a PerThreadCache in SSH queries to cache ProjectControl
instances inside DefaultPermissionBackend#ForProject so that
its performance is comparable to REST where the aforementioned
cache is already used.
This cache helps improve performance of the change visibility
check, which is done as a post-filter on the results obtained
from an index query.
Release-Notes: SSH query performance is improved
Change-Id: Ia0b5d8c10db4f6631c59af5d2e6631d7ca4d1ea0
|
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | | |
Obtaining the commit message is a costly operation as the commit
data has to be loaded. So, showing it even when --commit-message
is not provided degrades the performance of SSH queries.
Bug: Issue 15941
Release-Notes: SSH queries are fixed to show commit-message only when --commit-message is provided
Change-Id: I45cb478364535f1b9670cf60e3892898c2a46089
|
|\ \ \ \ \ \
| | | | | | |
| | | | | | |
| | | | | | | |
into stable-3.3
|
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | | |
Currently, we set Integer.MAX_VAL as the effective limit for an index
query processor when "--no-limit" option is specified. This throws
following exceptions on ElasticSearch index backend:
Result window is too large, from + size must be less
than or equal to: [2147483647] but was [2147501664]
Modify it make use of current index size + some extra buffer so that
ES server-side memory use is appropriately sized. Also add the API
in RepoSequence to return the last returned id.
Change-Id: I521b55a7c14c3780d1a91eb102cd56b1122e2061
Release-Notes: Fixed ES 'Result window is too large' error with --no-limit query option
|
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | | |
Hash tags are already available in ChangeData, use it, instead of
unnecessarily loading change notes to obtain them.
Bug: Issue 15942
Release-Notes: skip
Change-Id: Iad00c43adfe303c8e10777ce4744cb0709d9be30
|
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | | |
Release-Notes: skip
Change-Id: Id5f1ac35ed4f3f50b556bc3b3ec3712e2930ff94
|
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | | |
Release-Notes: skip
Change-Id: I1e4c75316601de11c9dc7512643b3a4801fb2035
|
|\ \ \ \ \ \ \
| |/ / / / / /
|/| / / / / /
| |/ / / / /
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | | |
* stable-3.2:
Set PerThreadCache as readonly after creating a new patch-set
Set PerThreadCache as readonly when formatting change e-mails
Set PerThreadCache as readonly when formatting change JSON
Set PerThreadCache as readonly after deleting a change
Set PerThreadCache as readonly after abandoning a change
Set PerThreadCache as readonly after merging a change
Set PerThreadCache as readonly after posting review comments
Introduce unloaders on PerThreadCache entries
RepoRefCache: Hold a reference to the refDatabase with ref counting
Remove use of RefCache in ChangeNotes
Revert "Cache change /meta ref SHA1 for each REST API request"
Cache change /meta ref SHA1 for each change indexing task
Release-Notes: skip
Change-Id: Ic3981c1fad0a1f8c232a72d858b8521d2407c0ff
|