[
https://issues.apache.org/jira/browse/COUCHDB-993?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12975817#action_12975817
]
Sebastian Cohnen commented on COUCHDB-993:
------------------------------------------
I cannot reproduce this error locally anymore. However I still see some odd
peaks in resource usage (memory/cpu). For now it looks okay, but I already said
to Paul on IRC couple of days ago that I'm going to perform some more extensive
testing and profiling to figure out where and - maybe with your help - why
these occur. I'm not sure if these usage patterns are expected or not, but I
hope that there is some more room for improvement here.
> Replication is crashing when changes feed was consumed
> ------------------------------------------------------
>
> Key: COUCHDB-993
> URL: https://issues.apache.org/jira/browse/COUCHDB-993
> Project: CouchDB
> Issue Type: Bug
> Components: Replication
> Affects Versions: 1.0.1, 1.0.2
> Reporter: Sebastian Cohnen
> Fix For: 1.0.2, 1.1
>
> Attachments: couchdb-993.patch
>
>
> Yesterday I hit a bug, where pull replication is dying which is resulting in
> a {exit,{json_encode,{bad_term,<0.133.0>}}} error (CouchDB is trying to
> encode a PID into JSON).
> Adam and Paul had a look at this issue yesterday and they found the
> underlying issue: There was a missing clause catching the exit message when
> the changes feed was consumed and ibrowse closes the HTTP connection for that
> feed.
> Adam wrote a quick patch yesterday, which I'll append here too (applies
> cleanly to 1.0.x at time of writing).
> (Sorry for any inaccuracy, I only understood the issue partially)
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
