incubator-couchdb-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Paul Davis <>
Subject Re: Errors in couchdb, happens everyday at specific time.
Date Wed, 26 Aug 2009 23:44:57 GMT
On Wed, Aug 26, 2009 at 7:25 PM, Niket Patel<> wrote:
> On Aug 27, 2009, at 4:40 AM, Paul Davis wrote:
>> Very interesting. Since it runs fine the rest of the time I'll have to
>> assume that you've got something running that is either changing file
>> permissions or locking files (eaccess points at either the files
>> changing permissions, or a directory in their path). I notice the
>> timestamps are from 11PM, is it  possible you've got a backup script
>> running that does something funky to the FS while it copies?
> I use EBS on EC2, and as of now I'm taking snapshots manually. I'm looking
> into crontab
> * This started after upgrade to 0.9.1
> * Happens at very specific time 4.00PM PST
> * This errors are logged when I start couchdb again, not when it died (its
> silent when died with log level info)
> * After start, it runs fine for next 24 hours
> data is safe ? ( This is my biggest concern now )

I'd be quite surprised to find that it was a 0.9.1 bug that introduced
errors in the code_server process at 4PM PST. I doubt that sort of bug
would go unnoticed  for this long :) Granted it could just be masking
the real cause, so who's to say?

Your biggest fear for data loss would be when clients can't connect to
the server because its borked. Once the data gets into couch assuming
you're ensuring a full commit then it should be peachy keen. But I
make no promises. This is the first time I've ever seen error messages
like those so for all I know you could have magic gnomes living in the
server rack.

Paul Davis

View raw message