[
https://issues.apache.org/jira/browse/KAFKA-420?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Jun Rao updated KAFKA-420:
--------------------------
Attachment: kafka-420_v1.patch
Attach patch v1.
This is a simpler fix. Instead of waiting for data flushed to disk, this patch
just advances the HW after the data is in memory. If a user cares about
durability, he can always increase the replication factor.
Patched an existing unit test to expose the problem. Patched 3 places where we
may need to increment the HW: (1) new produce requests coming to the leader;
(2) ISR shrinks; (3) a replica becomes the leader.
> maintain HW correctly with only 1 replica in ISR
> ------------------------------------------------
>
> Key: KAFKA-420
> URL: https://issues.apache.org/jira/browse/KAFKA-420
> Project: Kafka
> Issue Type: Bug
> Components: core
> Affects Versions: 0.8
> Reporter: Jun Rao
> Labels: bugs
> Attachments: kafka-420_v1.patch
>
> Original Estimate: 72h
> Remaining Estimate: 72h
>
> Currently, the HW maintenance logic is only triggered when handling fetch
> requests from the follower. As a result, if the ISR has only 1 replica, the
> HW won't be incremented since there is no request from the follower to
> trigger the maintenance logic.
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators:
https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa
For more information on JIRA, see: http://www.atlassian.com/software/jira
