[
https://issues.apache.org/jira/browse/IGNITE-24262?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17916364#comment-17916364
]
Philipp Shergalis commented on IGNITE-24262:
--------------------------------------------
Before creating a scheduler we wait for {{metaStorageSvcFut and
}}{{metaStorageConfigurationFuture}} to complete. So if these futures are
completed after node loses leadership, we will try to stop scheduler before it
was created. Reproduced this scenario in the same integration test by adding
delay in these futures and forcing a node to lose leadership and become leader
again. Also created a unit test reproducing this case.
Proof from original log: # *3344 became a leader*
2131 [15:26:11] : [:ignite-catalog-compaction:integrationTest]
[2025-01-13T11:26:11,524][INFO
][%icccct_n_3344%Raft-Group-Client-8][MetaStorageLeaderElectionListener] Node
has been elected as the leader (and it wasn't previous leader), so starting
doing secondary duties
*2. 3344 Tries to stop scheduler*
3843 [15:26:15] : [:ignite-catalog-compaction:integrationTest]
[2025-01-13T11:26:15,533][INFO
][%icccct_n_3344%JRaft-Request-Processor-10][MetaStorageLeaderElectionListener]
Node has lost the leadership, stopping doing secondary duties
*3.* *component start completes -> {{metaStorageSvcFut}} completes sometime
after that*
[15:26:18] : [:ignite-catalog-compaction:integrationTest]
[2025-01-13T11:26:18,187][INFO ][%icccct_n_3344%join-0][IgniteImpl] Components
started, performing recovery
*4. 3344 becomes a leader and tries to create scheduler - but first scheduler
was not stopped.*
4155 [15:26:19] : [:ignite-catalog-compaction:integrationTest]
[2025-01-13T11:26:19,748][INFO
][%icccct_n_3344%JRaft-Request-Processor-9][MetaStorageLeaderElectionListener]
Node has been elected as the leader (and it wasn't previous leader), so
starting doing secondary duties 4089 [15:26:19]W:
[:ignite-catalog-compaction:integrationTest]
[2025-01-13T11:26:19,749][ERROR][%icccct_n_3344%JRaft-Request-Processor-9][MetaStorageLeaderElectionListener]
Unable to start Idle Safe Time scheduler
> AssertionError in ClusterTimeImpl.startSafeTimeScheduler
> --------------------------------------------------------
>
> Key: IGNITE-24262
> URL: https://issues.apache.org/jira/browse/IGNITE-24262
> Project: Ignite
> Issue Type: Bug
> Reporter: Roman Puchkovskiy
> Assignee: Philipp Shergalis
> Priority: Major
> Labels: ignite-3
> Attachments: _Integration_Tests_Run_All_Other_31480.log.zip
>
> Time Spent: 10m
> Remaining Estimate: 0h
>
> AssertionError was thrown by ClusterTimeImpl.startSafeTimeScheduler() in
> ItCatalogCompactionCoordinatorChoosingTest, see the attached log
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
