[
https://issues.apache.org/jira/browse/KAFKA-17665?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17885970#comment-17885970
]
Satish Duggana commented on KAFKA-17665:
----------------------------------------
>What happen to segments that have the state COPY_SEGMENT_STARTED between
>server restarts?
The required segments will be started copying from the new leader. So, there
will not be any data loss but there can be duplicated segments in some
scenarios where the segment is successfully copied but it could not publish
COPY_SEGMENT_FINISHED metadata update.
> Kafka Tiered Storage retry back off is not used
> -----------------------------------------------
>
> Key: KAFKA-17665
> URL: https://issues.apache.org/jira/browse/KAFKA-17665
> Project: Kafka
> Issue Type: Bug
> Components: core, Tiered-Storage
> Reporter: Jorge Machado
> Priority: Minor
>
> Hi, we are evaluating the Kafka Tiered Storage. We already see KAFKA-17062
> which under 3.9 we could have data loss. I was looking at the code and also
> the KIP-405.
>
> On the KIP it is mentioned that if a copy fails it will be retried with a
> backoff propagation defined on
> REMOTE_LOG_MANAGER_TASK_RETRY_BACK_OFF_MS_PROP. The issue is the method
> remoteLogManagerTaskRetryBackoffMs() is not used in any place of the code. So
> I think this parameters make no effect currently. Is there any retry of
> failed segments ? Until now I just saw logging the error.
> Question from me: What happen to segments that have the state
> COPY_SEGMENT_STARTED between server restarts?
>
>
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
