[jira] [Commented] (KAFKA-12373) Improve KafkaRaftClient handling of graceful shutdown
[ https://issues.apache.org/jira/browse/KAFKA-12373?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17291449#comment-17291449 ] Sagar Rao commented on KAFKA-12373: --- hi [~jsancio], can i take this one up? > Improve KafkaRaftClient handling of graceful shutdown > - > > Key: KAFKA-12373 > URL: https://issues.apache.org/jira/browse/KAFKA-12373 > Project: Kafka > Issue Type: Sub-task > Components: replication >Reporter: Jose Armando Garcia Sancio >Priority: Major > > The current implementation simply closes the metrics group when it is closed. > When closing the KafkaRaftClient that is the leader it should perform at > least the following steps: > # Stop accepting new schedule append operations > # Append to the log the batches currently in the BatchAccumulator > # Wait with a timeout for the high-watermark to reach the LEO > # Cooperatively resign as leader from the quorum -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (KAFKA-12373) Improve KafkaRaftClient handling of graceful shutdown
[ https://issues.apache.org/jira/browse/KAFKA-12373?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17299975#comment-17299975 ] Jason Gustafson commented on KAFKA-12373: - [~sagarrao] Please feel free to pick up tasks if no one is currently assigned. Thanks for contributing! > Improve KafkaRaftClient handling of graceful shutdown > - > > Key: KAFKA-12373 > URL: https://issues.apache.org/jira/browse/KAFKA-12373 > Project: Kafka > Issue Type: Sub-task > Components: replication >Reporter: Jose Armando Garcia Sancio >Priority: Major > > The current implementation simply closes the metrics group when it is closed. > When closing the KafkaRaftClient that is the leader it should perform at > least the following steps: > # Stop accepting new schedule append operations > # Append to the log the batches currently in the BatchAccumulator > # Wait with a timeout for the high-watermark to reach the LEO > # Cooperatively resign as leader from the quorum -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (KAFKA-12373) Improve KafkaRaftClient handling of graceful shutdown
[ https://issues.apache.org/jira/browse/KAFKA-12373?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17300129#comment-17300129 ] Sagar Rao commented on KAFKA-12373: --- Thanks [~hachikuji], assigned it to myself > Improve KafkaRaftClient handling of graceful shutdown > - > > Key: KAFKA-12373 > URL: https://issues.apache.org/jira/browse/KAFKA-12373 > Project: Kafka > Issue Type: Sub-task > Components: replication >Reporter: Jose Armando Garcia Sancio >Assignee: Sagar Rao >Priority: Major > > The current implementation simply closes the metrics group when it is closed. > When closing the KafkaRaftClient that is the leader it should perform at > least the following steps: > # Stop accepting new schedule append operations > # Append to the log the batches currently in the BatchAccumulator > # Wait with a timeout for the high-watermark to reach the LEO > # Cooperatively resign as leader from the quorum -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (KAFKA-12373) Improve KafkaRaftClient handling of graceful shutdown
[ https://issues.apache.org/jira/browse/KAFKA-12373?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17309384#comment-17309384 ] Sagar Rao commented on KAFKA-12373: --- hi [~jsancio], I have started looking at this and had a question: There is already a mechanism for GracefulShutdown and then there areĀ close semantics in the KafkaRaftClient. In your Ticket title, you talk about graceful shutdown but in the description you say only metrics groups are closed. So, this behaviour on shutdown that you have described, needs to be added in shutdown/close or both? The reason i ask is that in the point #4 mentioned by you indicates that you want to trigger a Graceful shutdown, while in the javadocs for shutdown method in RaftClient, it is mentioned that it shouldn't be invoked in case of any exceptions. Effectively, all shutdown does is trigger a GracefulShutdown, so it is slightly confusing for me. > Improve KafkaRaftClient handling of graceful shutdown > - > > Key: KAFKA-12373 > URL: https://issues.apache.org/jira/browse/KAFKA-12373 > Project: Kafka > Issue Type: Sub-task > Components: replication >Reporter: Jose Armando Garcia Sancio >Assignee: Sagar Rao >Priority: Major > > The current implementation simply closes the metrics group when it is closed. > When closing the KafkaRaftClient that is the leader it should perform at > least the following steps: > # Stop accepting new schedule append operations > # Append to the log the batches currently in the BatchAccumulator > # Wait with a timeout for the high-watermark to reach the LEO > # Cooperatively resign as leader from the quorum -- This message was sent by Atlassian Jira (v8.3.4#803005)