You are viewing a plain text version of this content. The canonical link for it is here.
Posted to jira@kafka.apache.org by "Matthias J. Sax (Jira)" <ji...@apache.org> on 2025/07/19 00:21:00 UTC

[jira] [Updated] (KAFKA-19521) Fix consumption and leaving group when source topic is deleted

     [ https://issues.apache.org/jira/browse/KAFKA-19521?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Matthias J. Sax updated KAFKA-19521:
------------------------------------
    Flags:   (was: Patch)

> Fix consumption and leaving group when source topic is deleted
> --------------------------------------------------------------
>
>                 Key: KAFKA-19521
>                 URL: https://issues.apache.org/jira/browse/KAFKA-19521
>             Project: Kafka
>          Issue Type: Improvement
>            Reporter: travis
>            Priority: Major
>   Original Estimate: 168h
>  Remaining Estimate: 168h
>
> h2. Issue Description
> In the integration test `HandlingSourceTopicDeletionIntegrationTest`, we observe significant differences in application exit time between old and new protocols when source topic is deleted: * Old Protocol: ~10 seconds
>  * New Protocol: ~100 seconds
> h2. Observed Behavior
>  # State Transition Discrepancy:
>  ## One stream transitions to ERROR state quickly
>  ## Another stream remains in PENDING_ERROR state for an extended period
>  # Critical Log Evidence:
>  ** 
> {code:java}
> Leader for partition inputTopic-0 is unknown for fetching offset -1 Building ListOffsets request for partitions {inputTopic-0=-1}{code}
> {code:java}
> Consumer attempted to leave the group but couldn't complete it within 30000 ms. It will proceed to close.{code}
>  # Timing Analysis:
>  * 
>  ** 60s: AsyncConsumer continues attempting to consume while leader is unknown
>  ** 30s: Failed attempts to leave consumer group
>  ** Total: ~100s vs ~10s in old protocol
> h2. Root Cause Analysis
>  # Extended Polling Phase (60s):
>  ** AsyncConsumer continues polling despite unknown leader
>  ** No early termination mechanism when topic is deleted
>  ** Waits for full timeout before shutdown
>  # Group Leave Issues (30s):
>  ** New protocol struggles with group leave operation
>  ** Requires full timeout before forced close
>  ** Potential coordination issues in new protocol implementation
> h2. Related Code
>  * 
> {noformat}
> HandlingSourceTopicDeletionIntegrationTest{noformat}
>  * AsyncConsumer implementation
>  * New protocol group management code



--
This message was sent by Atlassian Jira
(v8.20.10#820010)