You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@flink.apache.org by "Niklas Semmler (Jira)" <ji...@apache.org> on 2022/03/09 10:59:00 UTC
[jira] [Comment Edited] (FLINK-25235) Re-enable ZooKeeperLeaderElectionITCase#testJobExecutionOnClusterWithLeaderChange
[ https://issues.apache.org/jira/browse/FLINK-25235?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17500246#comment-17500246 ]
Niklas Semmler edited comment on FLINK-25235 at 3/9/22, 10:58 AM:
------------------------------------------------------------------
Turns out this is more complicated than previously thought.
The test uses MiniCluster which currently assumes the use of a single High Availability services object and passes this class onto the multiple TaskManagers etc. In this test we create multiple JobManager and pass the same HA services to each JobManager. The test fails, when the second JobManager registers to the same HA services object as it expects only one.
This problem is restricted to the MiniCluster environment and more specifically, as we only use multiple HA services in tests, the TestingMiniCluster. However, these kind of tests add enough value to warrant changing the HA services object. [~dmvk], [~mapohl] and I discussed various options of multiplexing HA services to work with multiple JobManagers. For now we decided on adding an additional id to the HA services to multiplex the listeners in DefaultMultipleComponentLeaderElectionService.
was (Author: JIRAUSER281719):
Turns out this is more complicated than previously thought.
The test uses MiniCluster which currently assumes the use of a single High Availability services object and passes this class onto the multiple TaskManagers etc. In this test we create multiple JobManager and pass the same HA services to each JobManager. The test fails, when the second JobManager registers to the same HA services object as it expects only one.
This problem is restricted to the MiniCluster environment and more specifically, as we only use multiple HA services in tests, a test environment. However, these kind of tests add enough value to warrant changing the HA services object. [~dmvk], [~mapohl] and I discussed various options of multiplexing HA services to work with multiple JobManagers. For now we decided on adding an additional id to the HA services to multiplex the listeners in DefaultMultipleComponentLeaderElectionService.
> Re-enable ZooKeeperLeaderElectionITCase#testJobExecutionOnClusterWithLeaderChange
> ---------------------------------------------------------------------------------
>
> Key: FLINK-25235
> URL: https://issues.apache.org/jira/browse/FLINK-25235
> Project: Flink
> Issue Type: Technical Debt
> Components: Runtime / Coordination
> Reporter: David Morávek
> Assignee: Niklas Semmler
> Priority: Critical
> Labels: pull-request-available
> Fix For: 1.15.0
>
>
--
This message was sent by Atlassian Jira
(v8.20.1#820001)