[9.0] Fix deadlock in ThreadPoolMergeScheduler when a failing merge closes the IndexWriter (#134656) #135175
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
…the IndexWriter (elastic#134656) This change fixes a bug that causes a deadlock in the thread pool merge scheduler when a merge fails due to a tragic event. The deadlock occurs because Lucene aborts running merges when failing with a tragic event and then waits for them to complete. But those "running" merges might in fact be waiting in the Elasticsearch's thread pool merge scheduler tasks queue, or they might be waiting in the backlogged merge tasks queue because the per-shard concurrent merges count limit has been reached, or they might simply be waiting for enough disk space to be executed. In which cases the merge thread that is failing waits indefinitely. The proposed fix in this change uses the merge thread that is failing due to a tragic event to abort all other enqueued and backlogged merge tasks of the same shard, before pursuing with the closing of the IndexWriter. This way Lucene won't have to wait for any running merges as they would have all be aborted upfront. Relates ES-12664
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
This change fixes a bug that causes a deadlock in the thread pool merge scheduler when a merge fails due to a tragic event.
The deadlock occurs because Lucene aborts running merges when failing with a tragic event and then waits for them to complete. But those "running" merges might in fact be waiting in the Elasticsearch's thread pool merge scheduler tasks queue, or they might be waiting in the backlogged merge tasks queue because the per-shard concurrent merges count limit has been reached, or they might simply be waiting for enough disk space to be executed. In which cases the merge thread that is failing waits indefinitely.
The proposed fix in this change uses the merge thread that is failing due to a tragic event to abort all other enqueued and backlogged merge tasks of the same shard, before pursuing with the closing of the IndexWriter. This way Lucene won't have to wait for any running merges as they would have all be aborted upfront.
Backport of the #134656 for 9.0.8
Relates ES-12664