Caideyipi opened a new pull request, #19036:
URL: https://github.com/apache/iotdb/pull/19036

   ## Description
   
   Consensus subscription polling ignores the receiver's remaining timeout when 
polling region queues. Each dequeue attempt can wait for the full configured 
blocking interval, and acquiring the queue read lock has no timeout. With 
multiple regions or concurrent queue lifecycle operations, these waits can 
accumulate beyond the poll request's budget and cause client read timeouts.
   
   Seed one `PollTimer` from the receiver's remaining time and share it across 
topics, region queues, and dequeue retries. Limit both read-lock acquisition 
and blocking dequeue waits by the remaining budget, and stop starting new queue 
polls after expiration. Events already collected are still returned. Existing 
ownership, fallback, payload-limit, and acknowledgement handling is preserved.
   
   Runtime activation, WAL I/O, and response serialization retain their 
existing behavior and can still take additional time. This change does not 
address RPC buffer allocation retries under memory pressure.
   
   ## Validation
   
   - `mvn spotless:apply -P with-integration-tests -pl iotdb-core/datanode`
   - Clean DataNode dependency-closure build and 70 passing unit tests across 
`ConsensusSubscriptionBrokerPollTimeoutTest`, 
`ConsensusPrefetchingQueuePollTimeoutTest`, 
`ConsensusSubscriptionBrokerPayloadLimitTest`, 
`ConsensusSubscriptionBrokerOwnershipTest`, and 
`ConsensusPrefetchingQueueTest`. The seven new tests cover shared/decreasing 
budgets, receiver propagation, expiration, retained results, and lock 
contention.
   - Clean DataNode dependency-closure `test-compile` with `-P with-zh-locale 
-DskipTests`.
   - The same 70 tests also passed with `with-zh-locale` after strengthening 
the nonempty-queue/zero-budget fixture. DataNode was rebuilt from an empty 
target directory; unchanged upstream outputs were reused from the immediately 
preceding clean build with the same source, profile, and toolchain.
   
   Validation used local JDK 17 after remote repository connections timed out. 
The successful unit-test run used a 2 GB Maven heap and serial execution; test 
JVMs used a 64 MB initial heap, 1 GB maximum heap, and Serial GC. TsFile was 
built from clean source at `snapshot/2.5.1-261023` 
(`8f665d113dbb5257b901c528c31c8bc24da2a018`). No cluster load test was run.
   


-- 
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]

Reply via email to