[
https://issues.apache.org/jira/browse/BOOKKEEPER-401?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13454115#comment-13454115
]
Aniruddha commented on BOOKKEEPER-401:
--------------------------------------
Ah, you're right. However, this now gives us another case. Suppose we timeout
and invoke handleBookieFailure. While the metadata is being updated in memory
by one thread, we get a response from the bookie and ack the entry. We have now
acked an entry written to the old bookie.
Also, this would make the code more maintainable.
> LedgerHandle should not ack entries till an ensemble change has been
> persisted to Zookeeper
> -------------------------------------------------------------------------------------------
>
> Key: BOOKKEEPER-401
> URL: https://issues.apache.org/jira/browse/BOOKKEEPER-401
> Project: Bookkeeper
> Issue Type: Bug
> Components: bookkeeper-client
> Reporter: Aniruddha
>
> Consider the following case:
> Ensemble is A, B, C
> It changes to A, B, D locally and an update to write to ZK is in progress.
> We get a write request that writes to this ensemble, receives responses and
> acks back a success.
> The ensemble write to zk fails. We've acked an add for which there is no
> information in zookeeper.
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira
