incubator-couchdb-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From James Marca <jma...@translab.its.uci.edu>
Subject Couch 1.1.0 crashing on replication
Date Fri, 15 Jul 2011 03:27:48 GMT
Hi all,

Today I finally implemented an idea I'd had about doing some local
processing and then replicating documents to a central database.  I've
never used replication this intensively, and for the first time I'm
seeing CouchDB crash repeatedly.  

I'm calling replicate from node.js.  Both source and target are
running 1.1.0.  I'm running data processing on two machines and
replicating to a third.  I'm also running other couchdb tasks at the
same time on that third machine...mostly gets and bulk doc calls.

I can't say I've seen any pattern to the crashes, other than it
happens after I process just a handful of databases. The RAM usage
does creep up before the crash,  but it certainly doesn't trigger the
OOM killer.  The individual documents are large-ish.  Each database
can have at most 365 docs (one per day) and each database, when
compacted, runs to about 1.2GB.

I've seen the crash both with push and pull replication, with the
instigator of the replication (the pusher or the puller) being the one
that crashes.

Needless to say, this is irritating, as I want to just relax...

I'm aware this is a horrible bug report, but I want to get some
direction before I dump everything I know about the process.  

If anybody has any ideas on what I might look for (or if there are
known open issues) perhaps I can try to get the crash to happen "on
command" and then track down more information.  I have the logging set
to info, and there are no messages in the log file to indicate the
reason for the crash.  I hesitate to crank up the logging output
without a clear idea that I might be triggering the crash, as I really
just want to get my work done.

Thanks in advance for any help,

James Marca


Mime
View raw message