Dinesh Bhat has posted comments on this change.
Change subject: KUDU-1500: Fix the data race during
Patch Set 2:
TFTR Dan, please see responses inline, I would like to follow up on this
discussion on JIRA page, adding todd and mpercy in the loop there.
Line 11: the partition schema is accessed in an unguarded manner from another
> This is pretty confusing to me, because once a tablet is created, the parti
This follows the remote bootstrap(now tablet copy) path and there are at least
couple of use cases which fall under this. In the current situation, the leader
detects a failed tablet replica(perhaps via reporting? me not sure) and tries
to resurrect the on disk tablet metadata for that replica which is followed by
in-memory updates too. The attributes partition and schema are hanging off the
tablet metadata structure and hence inevitably become part of resurrection
path. We took an easier approach to this problem, although I agree with you
that it may need some more thoughts to elegantly solve this. I would like to
take the discussion in JIRA because it's easier to see the history of these
discusssions on JIRA page than on gerritt review page IMO.
Line 177: attrs["partition"] =
> if this is just a formatting change, please leave it the way it was.
Sounds good, it looks like I touched some of these lines, but forgot to revert
them back to original format when I reverted them later.
To view, visit http://gerrit.cloudera.org:8080/3823
To unsubscribe, visit http://gerrit.cloudera.org:8080/settings
Gerrit-Owner: Dinesh Bhat <din...@cloudera.com>
Gerrit-Reviewer: Dan Burkert <d...@cloudera.com>
Gerrit-Reviewer: Dinesh Bhat <din...@cloudera.com>
Gerrit-Reviewer: Kudu Jenkins
Gerrit-Reviewer: Mike Percy <mpe...@apache.org>
Gerrit-Reviewer: Todd Lipcon <t...@apache.org>