[ https://issues.apache.org/jira/browse/KAFKA-9656?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Jason Gustafson updated KAFKA-9656: ----------------------------------- Description: In KAFKA-7296, we fixed a bug which causes the producer to enter a fatal state when the COORDINATOR_LOADING_IN_PROGRESS error is received. The impact of this bug in streams was that the application would crash. Generally we want users to upgrade to a later client version, but in some cases, this takes a long time. I am suggesting here that we revert the behavior change on the broker for older versions of TxnOffsetCommit. For versions older than 2 (which was introduced in 2.3), rather than returning COORDINATOR_LOADING_IN_PROGRESS, we can return COORDINATOR_NOT_AVAILABLE. (was: In KAFKA-7296, we fixed a bug which causes the producer to enter a fatal state when the COORDINATOR_LOADING_IN_PROGRESS error is received. The impact of this bug in streams was that the application would crash. Generally we want users to upgrade to a later client version, but in some cases, this takes a long time. I am suggesting here that we revert this behavior for older versions of TxnOffsetCommit. For versions older than 2 (which was introduced in 2.3), rather than returning COORDINATOR_LOADING_IN_PROGRESS, we can return COORDINATOR_NOT_AVAILABLE.) > TxnOffsetCommit should not return COORDINATOR_LOADING error for old request > versions > ------------------------------------------------------------------------------------ > > Key: KAFKA-9656 > URL: https://issues.apache.org/jira/browse/KAFKA-9656 > Project: Kafka > Issue Type: Bug > Reporter: Jason Gustafson > Priority: Major > > In KAFKA-7296, we fixed a bug which causes the producer to enter a fatal > state when the COORDINATOR_LOADING_IN_PROGRESS error is received. The impact > of this bug in streams was that the application would crash. Generally we > want users to upgrade to a later client version, but in some cases, this > takes a long time. I am suggesting here that we revert the behavior change on > the broker for older versions of TxnOffsetCommit. For versions older than 2 > (which was introduced in 2.3), rather than returning > COORDINATOR_LOADING_IN_PROGRESS, we can return COORDINATOR_NOT_AVAILABLE. -- This message was sent by Atlassian Jira (v8.3.4#803005)