Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

[Bug] [Segment Replication] Fix isAheadOf logic for ReplicationCheckpoint comparison #4112

Conversation

dreamer-89
Copy link
Member

@dreamer-89 dreamer-89 commented Aug 3, 2022

Signed-off-by: Suraj Singh [email protected]

Description

Coming from #3988 which relies on ReplicationCheckpoint ordering for correctly identifying the furthest ahead replica. The existing ReplicationCheckpoint#isAheadOf is not correct as it uses || b/w segmentInfosVersion (1st) & PrimaryTerm(2nd) terms; instead of relying on 2nd term (PrimaryTerm field) ONLY WHEN first field (segmentInfosVersion) is equal.

This PR also fixes, the comparison order i.e. PrimaryTerm first followed by segmentInfosVersion; this is to prioritize active primary over stale.

Issues Related

#3988

Related

#4041

By submitting this pull request, I confirm that my contribution is made under the terms of the Apache 2.0 license.
For more information on following Developer Certificate of Origin and signing off your commits, please check here.

@dreamer-89 dreamer-89 requested review from a team and reta as code owners August 3, 2022 19:04
@github-actions
Copy link
Contributor

github-actions bot commented Aug 3, 2022

Gradle Check (Jenkins) Run Completed with:

@dreamer-89
Copy link
Member Author

> Task :distribution:bwc:staged:buildBwcLinuxTar FAILED

> Task :distribution:bwc:minor:buildBwcLinuxTar
 [2.3.0] > Task :buildSrc:pluginUnderTestMetadata FAILED
 [2.3.0] 
 [2.3.0] 11 actionable tasks: 11 executed
 [2.3.0] FAILURE: Build failed with an exception.
 [2.3.0] 
 [2.3.0] * What went wrong:
 [2.3.0] Execution failed for task ':buildSrc:pluginUnderTestMetadata'.
 [2.3.0] > Timeout waiting to lock file hash cache (/home/ubuntu/.gradle/caches/7.5/fileHashes). It is currently in use by another Gradle instance.
 [2.3.0]   Owner PID: 28733
 [2.3.0]   Our PID: 32188
 [2.3.0]   Owner Operation: 
 [2.3.0]   Our operation: 
 [2.3.0]   Lock file: /home/ubuntu/.gradle/caches/7.5/fileHashes/fileHashes.lock
 [2.3.0] 
 [2.3.0] * Try:
 [2.3.0] > Run with --stacktrace option to get the stack trace.
 [2.3.0] > Run with --info or --debug option to get more log output.
 [2.3.0] 
 [2.3.0] * Get more help at https://help.gradle.org
 [2.3.0] 
 [2.3.0] BUILD FAILED in 2m 4s
Output for /var/jenkins/workspace/gradle-check/search/distribution/bwc/minor/build/bwc/checkout-2.x/gradlew:

> Task :distribution:bwc:minor:buildBwcLinuxTar FAILED

@github-actions
Copy link
Contributor

github-actions bot commented Aug 3, 2022

Gradle Check (Jenkins) Run Completed with:

@codecov-commenter
Copy link

Codecov Report

Merging #4112 (c992024) into main (6f43dbc) will decrease coverage by 0.07%.
The diff coverage is 50.00%.

@@             Coverage Diff              @@
##               main    #4112      +/-   ##
============================================
- Coverage     70.61%   70.53%   -0.08%     
- Complexity    56983    57004      +21     
============================================
  Files          4595     4595              
  Lines        274234   274236       +2     
  Branches      40181    40183       +2     
============================================
- Hits         193638   193427     -211     
- Misses        64342    64564     +222     
+ Partials      16254    16245       -9     
Impacted Files Coverage Δ
...pensearch/common/settings/IndexScopedSettings.java 100.00% <ø> (ø)
.../java/org/opensearch/snapshots/RestoreService.java 55.37% <0.00%> (ø)
.../replication/checkpoint/ReplicationCheckpoint.java 53.48% <33.33%> (-0.18%) ⬇️
...org/opensearch/cluster/metadata/IndexMetadata.java 83.85% <100.00%> (ø)
.../main/java/org/opensearch/index/IndexSettings.java 86.35% <100.00%> (ø)
...a/org/opensearch/client/cluster/SniffModeInfo.java 0.00% <0.00%> (-58.83%) ⬇️
...a/org/opensearch/tasks/TaskCancelledException.java 50.00% <0.00%> (-50.00%) ⬇️
...cluster/coordination/PendingClusterStateStats.java 20.00% <0.00%> (-48.00%) ⬇️
.../org/opensearch/client/indices/AnalyzeRequest.java 31.00% <0.00%> (-45.00%) ⬇️
...rch/client/transport/NoNodeAvailableException.java 28.57% <0.00%> (-42.86%) ⬇️
... and 470 more

Help us with your feedback. Take ten seconds to tell us how you rate us.

@dreamer-89 dreamer-89 merged commit 7af8e37 into opensearch-project:main Aug 3, 2022
Rishikesh1159 pushed a commit to Rishikesh1159/OpenSearch that referenced this pull request Aug 17, 2022
Rishikesh1159 added a commit that referenced this pull request Aug 17, 2022
…aining segment replication changes (#4243)

* [Segment Replication] Checkpoint Replay on Replica Shard (#3658)

* Adding Latest Recevied checkpoint, replay checkpoint logic along with tests

Signed-off-by: Rishikesh1159 <[email protected]>

* [Segment Replication] Wire up segment replication with peer recovery and add ITs. (#3743)

* Add null check when computing max segment version.

With segment replication enabled it is possible Lucene does not set the SegmentInfos
min segment version, leaving the default value as null.

Signed-off-by: Marc Handalian <[email protected]>

* Update peer recovery to set the translogUUID of replicas to the UUID generated on the primary.

This change updates the UUID when the translog is created to the value stored in the passed segment userdata.
This is to ensure during failover scenarios that the replica can be promoted and not have a uuid mismatch with the value stored in user data.

Signed-off-by: Marc Handalian <[email protected]>

* Wire up Segment Replication under the feature flag.

This PR wires up segment replication and adds some initial integration tests.

Signed-off-by: Marc Handalian <[email protected]>

* Add test to ensure replicas use primary translog uuid with segrep.

Signed-off-by: Marc Handalian <[email protected]>

* Update SegmentReplicationIT to assert previous commit points are valid and SegmentInfos can be built.
Fix nitpicks in PR feedback.

Signed-off-by: Marc Handalian <[email protected]>

* Fix test with Assert.fail to include a message.

Signed-off-by: Marc Handalian <[email protected]>

* Disable shard idle with segment replication. (#4118)

This change disables shard idle when segment replication is enabled.
Primary shards will only push out new segments on refresh, we do not want to block this based on search behavior.
Replicas will only refresh on an externally provided SegmentInfos, so we do not want search requests to hang waiting for a refresh.

Signed-off-by: Marc Handalian <[email protected]>

* Fix isAheadOf logic for ReplicationCheckpoint comparison (#4112)

Signed-off-by: Suraj Singh <[email protected]>

* Fix waitUntil refresh policy for segrep enabled indices. (#4137)

Signed-off-by: Marc Handalian <[email protected]>

* Add IndexShard#getLatestReplicationCheckpoint behind segrep enable feature flag (#4163)

* Add IndexShard#getLatestReplicationCheckpoint behind segrep enable feature flag

Signed-off-by: Suraj Singh <[email protected]>

* Address review comment. Move tests to SegmentReplicationIndexShardTests

Signed-off-by: Suraj Singh <[email protected]>

* Add segrep enbaled index settings in TargetServiceTests, SourceHandlerTests

Signed-off-by: Suraj Singh <[email protected]>

* [Segment Replication] Fix OngoingSegmentReplications to key by allocation ID instead of DiscoveryNode. (#4182)

* Fix OngoingSegmentReplications to key by allocation ID instead of DiscoveryNode.

This change fixes segrep to work with multiple shards per node by keying ongoing replications on
allocation ID.  This also updates cancel methods to ensure state is properly cleared on shard cancel.

Signed-off-by: Marc Handalian <[email protected]>

* Clean up cancel methods.

Signed-off-by: Marc Handalian <[email protected]>

Signed-off-by: Marc Handalian <[email protected]>

* [Bug] [Segment Replication] Update store metadata recovery diff logic to ignore missing files causing exception (#4185)

* Update Store for segment replication dif

Signed-off-by: Poojita Raj <[email protected]>

Signed-off-by: Poojita Raj <[email protected]>

* Update recoveryDiff logic to ingore missing files causing exception on replica during copy

Signed-off-by: Suraj Singh <[email protected]>

* Address review comments

Signed-off-by: Suraj Singh <[email protected]>

Signed-off-by: Poojita Raj <[email protected]>
Signed-off-by: Suraj Singh <[email protected]>
Co-authored-by: Poojita Raj <[email protected]>

* [Segment Replication] Adding PrimaryMode check before publishing checkpoint and processing a received checkpoint. (#4157)

* Adding PrimaryMode check before publishing checkpoint.

Signed-off-by: Rishikesh1159 <[email protected]>

* Applying spotless check

Signed-off-by: Rishikesh1159 <[email protected]>

* Moving segrep specific tests to SegmentReplicationIndexShardTests.

Signed-off-by: Rishikesh1159 <[email protected]>

* Adding logic and tests for rejecting checkpoints if shard is in PrimaryMode.

Signed-off-by: Rishikesh1159 <[email protected]>

* Applying ./gradlew :server:spotlessApply.

Signed-off-by: Rishikesh1159 <[email protected]>

* Applying ./gradlew :server:spotlessApply

Signed-off-by: Rishikesh1159 <[email protected]>

* Changing log level to warn in shouldProcessCheckpoint() of IndexShard.java class.

Signed-off-by: Rishikesh1159 <[email protected]>

* Removing unnecessary lazy logging in shouldProcessCheckpoint().

Signed-off-by: Rishikesh1159 <[email protected]>

Signed-off-by: Rishikesh1159 <[email protected]>

* [Segment Replication] Wait for documents to replicate to replica shards (#4236)

* [Segment Replication] Add thread sleep to account for replica lag in delete operations test

Signed-off-by: Suraj Singh <[email protected]>

* Address review comments, assertBusy on doc count rather than sleep

Signed-off-by: Suraj Singh <[email protected]>

Signed-off-by: Suraj Singh <[email protected]>

* Segment Replication - Remove unnecessary call to markAllocationIdAsInSync. (#4224)

This PR Removes an unnecessary call to markAllocationIdAsInSync on the primary shard when replication events complete.
Recovery will manage this initial call.

Signed-off-by: Marc Handalian <[email protected]>

Signed-off-by: Marc Handalian <[email protected]>

* Segment Replication - Add additional unit tests for update & delete ops. (#4237)

* Segment Replication - Add additional unit tests for update & delete operations.

Signed-off-by: Marc Handalian <[email protected]>

* Fix spotless.

Signed-off-by: Marc Handalian <[email protected]>

Signed-off-by: Marc Handalian <[email protected]>

Signed-off-by: Rishikesh1159 <[email protected]>
Signed-off-by: Marc Handalian <[email protected]>
Signed-off-by: Suraj Singh <[email protected]>
Signed-off-by: Poojita Raj <[email protected]>
Co-authored-by: Marc Handalian <[email protected]>
Co-authored-by: Suraj Singh <[email protected]>
Co-authored-by: Poojita Raj <[email protected]>
@dreamer-89 dreamer-89 added the backport 2.x Backport to 2.x branch label Sep 2, 2022
@opensearch-trigger-bot
Copy link
Contributor

The backport to 2.x failed:

The process '/usr/bin/git' failed with exit code 1

To backport manually, run these commands in your terminal:

# Fetch latest updates from GitHub
git fetch
# Create a new working tree
git worktree add .worktrees/backport-2.x 2.x
# Navigate to the new working tree
cd .worktrees/backport-2.x
# Create a new branch
git switch --create backport/backport-4112-to-2.x
# Cherry-pick the merged commit of this pull request and resolve the conflicts
git cherry-pick -x --mainline 1 7af8e37f81c0a124712a8cf86dab973df26f906f
# Push it to GitHub
git push --set-upstream origin backport/backport-4112-to-2.x
# Go back to the original working tree
cd ../..
# Delete the working tree
git worktree remove .worktrees/backport-2.x

Then, create a pull request where the base branch is 2.x and the compare/head branch is backport/backport-4112-to-2.x.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
backport 2.x Backport to 2.x branch
Projects
None yet
Development

Successfully merging this pull request may close these issues.

4 participants