[
https://issues.apache.org/jira/browse/KAFKA-20720?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Lianet Magrans resolved KAFKA-20720.
------------------------------------
Resolution: Fixed
> Review level for client-side log on share session errors
> --------------------------------------------------------
>
> Key: KAFKA-20720
> URL: https://issues.apache.org/jira/browse/KAFKA-20720
> Project: Kafka
> Issue Type: Bug
> Reporter: Nilesh Kumar
> Assignee: Nilesh Kumar
> Priority: Major
>
> When share sessions are evicted or get out of sync (broker-managed), the
> share consumer gets {{{}SHARE_SESSION_NOT_FOUND{}}},
> {{{}INVALID_SHARE_SESSION_EPOCH{}}}, or {{SHARE_SESSION_LIMIT_REACHED}}
> logged at INFO in {{{}ShareSessionHandler{}}}, with an alarming message like
> {{{}"Node {} was unable to process the ShareFetch/ShareAcknowledge
> request"{}}}.
> * This is self-healing on the client (re-sends a full
> ShareFetch/ShareAcknowledge via {{{}nextCloseExistingAttemptNew(){}}}); no
> client-side action is possible.
> * Session cache is broker-managed; visibility (metrics, configs) lives on
> the broker.
> * Broker-side logs for epoch mismatch are already at DEBUG
> ({{{}SharePartitionManager{}}}, {{{}ShareSessionContext{}}}).
> Related: KAFKA-20713 (fetch equivalent, fixed in PR #22617).
> This Jira is to review the client-side log in {{ShareSessionHandler}}
> (consider DEBUG + {{{}isDebugEnabled{}}}, aligned with
> {{{}FetchSessionHandler{}}}, so apps keep relevant INFO without this noise).
>
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
