merlimat opened a new pull request, #25827:
URL: https://github.com/apache/pulsar/pull/25827
### Motivation
`CompactionConcurrencyTest.testDisableCompactionConcurrently` fires two
threads that both call `deleteSubscriptionAsync(\"__compaction\")` on the same
topic and asserts that **at least one** of them fails:
```java
assertTrue(f1.get().isCompletedExceptionally() ||
f2.get().isCompletedExceptionally());
```
That assertion has been flaky since the test was migrated from
`ProducerConsumerBase` to `SharedPulsarBaseTest` in #25381.
The original test relied on `MockZooKeeper.delay(...)` to inject a delay
into the cursor delete operation. This forced the two calls to overlap inside
`PersistentTopic#asyncDeleteCursorWithCleanCompactionLedger`'s `synchronized`
block — the second one then hit the `disablingCompaction.compareAndSet(false,
true) == false` branch and failed with `SubscriptionBusyException`.
`SharedPulsarBaseTest` uses an in-memory metadata store and there is no
equivalent delay hook. The first delete can complete (and remove the
subscription) before the second thread even reaches the broker. The second call
then runs through `asyncDeleteCursorWithClearDelayedMessage`, sees
`subscriptions.get(\"__compaction\") == null`, and returns success as a no-op:
```java
PersistentSubscription persistentSubscription =
subscriptions.get(subscriptionName);
if (persistentSubscription == null) {
log.warn().attr(...).log(\"Can't find subscription, skip delete
cursor\");
unsubscribeFuture.complete(null);
return;
}
```
Both futures complete successfully — perfectly valid behavior — and the
assertion fails.
Example failure:
https://scans.gradle.com/s/2odpjlsucvuly/tests/task/:pulsar-broker:test/details/org.apache.pulsar.broker.service.persistent.CompactionConcurrencyTest/testDisableCompactionConcurrently/2/output
### Modifications
Drop the \"at least one fails\" assertion — the broker handles either
outcome correctly (concurrent overlap → `SubscriptionBusyException`; sequential
→ no-op).
Keep the deadlock check (both futures complete in bounded time) and add a
verification that the compaction subscription ends up removed, which is the
actual invariant we care about.
### Verifying this change
This change is already covered by
`CompactionConcurrencyTest.testDisableCompactionConcurrently`. Locally I ran 5
times with `--rerun-tasks`; all passed.
### Does this pull request potentially affect one of the following parts:
- [ ] Dependencies (add or upgrade a dependency)
- [ ] The public API
- [ ] The schema
- [ ] The default values of configurations
- [ ] The threading model
- [ ] The binary protocol
- [ ] The REST endpoints
- [ ] The admin CLI options
- [ ] The metrics
- [ ] Anything that affects deployment
--
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]
