[
https://issues.apache.org/jira/browse/KUDU-495?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15260923#comment-15260923
]
Alex Feinberg commented on KUDU-495:
------------------------------------
Glad to see this! (Have been watching these issues a bit and read the updated
design doc, but haven't had the time to dive into the reviews.)
> Master should handle the case where we REPLICATE fails
> ------------------------------------------------------
>
> Key: KUDU-495
> URL: https://issues.apache.org/jira/browse/KUDU-495
> Project: Kudu
> Issue Type: Sub-task
> Components: master
> Affects Versions: M5, 0.7.0
> Reporter: Alex Feinberg
> Assignee: Adar Dembo
> Priority: Critical
> Fix For: 0.9.0
>
>
> Master should handle the case where the REPLICATE fails (which I discovered
> when running a transaction against a non-leader case) in SysTables.
> This is a bit tricky as for master we update the in-memory tables *first*,
> then commit the transaction.
> Only way would be able to do the following:
> Under the lock for the appropriate CopyOnWrite object--
> 1) Save the old version of the object.
> 2) Make the appropriate modification.
> 3) If REPLICATE fails, go back to the old version.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
