[
https://issues.apache.org/jira/browse/COUCHDB-2289?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14131437#comment-14131437
]
Gil Pedersen commented on COUCHDB-2289:
---------------------------------------
[~kxepal] Thanks for taking an interest.
By stall I mean that at some point the replicating client won't progress past
some sequence checkpoint. For instance it now shows "Checkpointed source
sequence 2253821, current source sequence 2295766, progress 98%". It will stay
indefinitely at "2253821" but will update current source sequence number when
it changes, ie. network communication is working. If I restart the replicating
instance, it will progress past the point that it stalled at, sometimes
stalling at another sequence number before reaching 100%.
Another note is that non-couchdb clients that use the _changes feed like
elasticsearch works fine.
I have not looked at tcpdumps yet.
> Replication stall
> -----------------
>
> Key: COUCHDB-2289
> URL: https://issues.apache.org/jira/browse/COUCHDB-2289
> Project: CouchDB
> Issue Type: Bug
> Security Level: public(Regular issues)
> Components: Replication
> Reporter: Gil Pedersen
>
> I have 2 couchdb instances in a local setup with one acting as the master,
> and the other performing a continuous pull replication using the _replicator
> db.
> This was initially working great but now I occasionally experience permanent
> stalls in the replication. I can get it running again by restarting the
> couchdb instance but it will eventually stall again.
> Gist with more info here: https://gist.github.com/kanongil/9788322
> I have also tried to create a 1.6.0 docker-based instance which also stalls
> (but at different sequence numbers).
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
