JAkutenshi commented on code in PR #4446:
URL: https://github.com/apache/ignite-3/pull/4446#discussion_r1871335037
##########
modules/catalog-compaction/src/integrationTest/java/org/apache/ignite/internal/catalog/compaction/ItCatalogCompactionTest.java:
##########
@@ -51,14 +51,13 @@
import org.apache.ignite.network.ClusterNode;
import org.apache.ignite.tx.Transaction;
import org.apache.ignite.tx.TransactionOptions;
-import org.junit.jupiter.api.BeforeAll;
import org.junit.jupiter.api.BeforeEach;
import org.junit.jupiter.api.Test;
/**
* Integration tests to verify catalog compaction.
*/
-class ItCatalogCompactionTest extends ClusterPerClassIntegrationTest {
+class ItCatalogCompactionTest extends ClusterPerTestIntegrationTest {
Review Comment:
Due to personal discussion there was found that the drastic reason relates
to [IGNITE-23753](https://issues.apache.org/jira/browse/IGNITE-23753).
If we'll add replica state on assertion at the revision `1fe7f5ff` we could
see then `ASSIGNED` state for already removed replicas. The problem is similar
to [IGNITE-23753](https://issues.apache.org/jira/browse/IGNITE-23753), but
reason differ: we have 2 places in `TableManager` where `stopReplica` raw calls
are done:
1. [IGNITE-23753](https://issues.apache.org/jira/browse/IGNITE-23753)'s
issue on Ignite node stopping process;
2. on drop table process: `onLwmChanged` -> `destroyTableLocally` ->
`stopAndDestroyPartition` -> `stopPartition` -> `stopReplica`.
In both cases we call `stopReplica` instead `weakStopReplica` with a
particular reason and replica's states handling. It leads to wrong states and
possible memory leak because `ASSIGNED` states shouldn't be deleted.
The issue should be investigated in
[IGNITE-23753](https://issues.apache.org/jira/browse/IGNITE-23753), for now
test class after 20 repeated runs is green locally due to no-op handling on
null-value replica future. But bug with states is confirmed.
--
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]
