fix: fix scheduler deadlock when reducing max_outstanding_requests_per_tenant #7188
+58
−1
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
…urs when a tenant's max_outstanding_requests_per_tenant limit is dynamically reduced via runtime configuration. When MaxOutstandingPerTenant is reduced while a user's FIFORequestQueue is full, the getOrAddQueue method attempts to migrate requests to a smaller queue. Previously, this loop blocked indefinitely when the new queue capacity was reached, causing the scheduler to freeze. The fix ensures the migration loop breaks when the new queue is full, effectively dropping excess requests instead of blocking. Signed-off-by: Kishore K G <[email protected]>
Signed-off-by: Kishore K G <[email protected]>
kishorekg1999
changed the title
Signed-off-by: Kishore K G <[email protected]>
Contributor
Author
|
@SungJin1212 can you please review this PR when you get a chance ? |
SungJin1212
reviewed
Jan 8, 2026
|
|
||
| // flush to new queue | ||
| for uq.queue.length() > 0 { | ||
| for (uq.queue.length() > 0) && (tmpQueue.length() < maxOutstanding) { |
Member
There was a problem hiding this comment.
can you add some comment to this line?
like
Suggested change
| for (uq.queue.length() > 0) && (tmpQueue.length() < maxOutstanding) { | |
| // If the new limit is lower than the current number of requests, | |
| // the excess requests (newest ones) will be dropped to prevent deadlocks. | |
| for (uq.queue.length() > 0) && (tmpQueue.length() < maxOutstanding) { |
SungJin1212
reviewed
Jan 8, 2026
Member
SungJin1212
left a comment
SungJin1212
left a comment
There was a problem hiding this comment.
The changes look good to me, I left one comment. Thanks for contribution!
Member
|
@kishorekg1999 |
Signed-off-by: Kishore K G <[email protected]>
Contributor
Author
|
Thanks for the quick response @SungJin1212 ! I've incorporated the comments. |
Signed-off-by: Kishore K G <[email protected]>
SungJin1212
reviewed
Jan 8, 2026
| * [FEATURE] You can specify "heap ballast" to reduce Go GC Churn #1489 | ||
| * [BUGFIX] HA Tracker no longer always makes a request to Consul/Etcd when a request is not from the active replica #1516 | ||
| * [BUGFIX] Queries are now correctly cancelled by the query-frontend #1508 | ||
| * [BUGFIX] If max_outstanding_requests_per_tenant value is updated to lesser value than the current number of requests in the queue, the excess requests (newest ones) will be dropped to prevent deadlocks. #7188 |
Member
There was a problem hiding this comment.
Can you move it to master / unreleased? and would be good to add target component like Scheduler: If max_outstanding_requests_per_tenant ~.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
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.
What this PR does:
This PR fixes a critical deadlock in the query scheduler that occurs when a tenant's max_outstanding_requests_per_tenant limit is dynamically reduced via runtime configuration.
When MaxOutstandingPerTenant is reduced while a user's FIFORequestQueue is full, the getOrAddQueue method attempts to migrate requests to a smaller queue. Previously, this loop blocked indefinitely when the new queue capacity was reached, causing the scheduler to freeze.
The fix ensures the migration loop breaks when the new queue is full, effectively dropping excess requests instead of blocking.
Checklist
CHANGELOG.mdupdated - the order of entries should be[CHANGE],[FEATURE],[ENHANCEMENT],[BUGFIX]