GJ100 opened a new pull request, #68178:
URL: https://github.com/apache/doris/pull/68178
### What problem does this PR solve?
Issue Number: close #68169
Problem Summary:
When a MySQL client terminates its connection abruptly (e.g., client-side
timeout elapsed while BE was slow or waiting for lock), FE's network layer
invokes ConnectPoolMgr.unregisterConnection(context).
However, unregisterConnection previously only removed the context from
connectionMap without cancelling the running query. This caused multiple severe
issues:
1. Timeout Checker Bypass: FE's TimeoutChecker (checkTimeout) only inspects
connections present in connectionMap.values(). Once removed, the connection is
never checked again, allowing the orphaned query to bypass query_timeout
entirely and run indefinitely.
2. Ghost Queries in active_queries: The query stayed RUNNING indefinitely in
QeProcessorImpl and information_schema.active_queries because the coordinator
was never finalized.
3. Workload Group Concurrency Starvation: The query never released its
QueueToken in WorkloadGroup. When all group slots were consumed by such
orphaned queries, all subsequent queries became permanently queued in
WAIT_IN_QUEUE until timing out.
Fix:
1. In ConnectPoolMgr.unregisterConnection(), asynchronously cancel running
queries via ctx.cancelQuery(..., false) without blocking the XNIO
close-listener thread, ensuring coordinator aborts BE fragment execution,
releases workload group QueueTokens, and unblocks waiting FE threads.
2. Added cancelQuery(Status, boolean needWaitCancelComplete) in
ConnectContext and defensive cancelQuery calls in
ConnectContext.killConnection() and cleanup().
3. Added unit tests in ConnectSchedulerTest and ConnectContextTest.
### Release note
Fix orphaned queries hanging in information_schema.active_queries and
leaking Workload Group concurrency slots when clients disconnect.
### Check List (For Author)
- Test
- [ ] Regression test
- [x] Unit Test
- [ ] Manual test (add detailed scripts or steps below)
- [ ] No need to test or manual test. Explain why:
- Behavior changed: No
- Does this need documentation: No
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
