[
https://issues.apache.org/jira/browse/KUDU-1042?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15705079#comment-15705079
]
Mike Percy commented on KUDU-1042:
----------------------------------
[~tlipcon] I am current a bit fuzzy on this one. We worked around the issue in
https://github.com/apache/kudu/commit/f9e0d4ae8b78d15337912db2f9e62a4e5fa06418
so I am not sure this is required anymore.
> Consider serializing timestamp assignment for non-tablet operations
> -------------------------------------------------------------------
>
> Key: KUDU-1042
> URL: https://issues.apache.org/jira/browse/KUDU-1042
> Project: Kudu
> Issue Type: Improvement
> Components: consensus
> Affects Versions: Feature Complete
> Reporter: Mike Percy
>
> We currently may have non-monotonic timestamp assignment to opids because
> write transaction prepare may run concurrently with change config operations.
> This is the cause of the bootstrap crasher seen in KUDU-1035.
> We don't think it's currently critical (as of public beta) to enforce this
> monotonicity, but we should revisit this when we put more thought into the
> ramifications as it relates to safe time.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
