incubator-couchdb-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Robert Newson <rnew...@apache.org>
Subject Re: unusable database in 1.0.1. steps to take?
Date Thu, 04 Aug 2011 16:34:49 GMT
Hm, yeah, that might not be practical.

Does the crash happen immediately on startup or do you need to query a
document or a view first?

B.

On 4 August 2011 17:07, Jason Konrad <jason@inboxq.com> wrote:
> The file system is ext3 and delayed_commits is set to false.
>
> I think its possible to share the file although its 65GB so I don't know how
> practical it would be.
> On Aug 4, 2011 2:01 AM, "Robert Newson" <rnewson@apache.org> wrote:
>> Hi Jason,
>>
>> What filesystem is this stored on? Are you running with
>> delayed_commits set to true or false?
>>
>> Finally, are you able to share the database file with the CouchDB
>> development team?
>>
>> B.
>>
>> On 4 August 2011 03:39, Jason Konrad <jason@inboxq.com> wrote:
>>> I have a database that has become unusable today. The only way I can
>>> get the couchdb server to function is to remove "thedb" which it
>>> doesn't like. I've attached a file with some of the log data as well
>>> as some errorr futon throws up. I'm looking for any ideas for next
>>> steps to try and "recover" this database.
>>>
>>> This is running on a CentOS5 system with erlang-R12B-5.12.el5 and
>>> couchdb-1.0.1-2.el5
>>>
>>> Let me give you a little context to how this came about. I have one
>>> database, not the one with problems, that has a high update rate on
>>> documents. Combined with smallish disks this requires compaction a
>>> couple times a week. The compactions have been going well but the data
>>> is not always freed up from the system after the compaction. In order
>>> to get the disk space to free up I restart couchdb. This process had
>>> been working for a few months but not today.
>>>
>>> On the restart of couchdb the first signs of something out of the
>>> ordinary was happening, two views were being completely recalulated.
>>> These calculations can take hours to finish but today that was not the
>>> case. After about 30 min the views stopped calculating and couchdb
>>> started throwing 500 response codes for everything no matter what
>>> database was being used ( there are 26 total, 1 bad one ). I tried
>>> restarting couchdb but all that would happen is the log file would get
>>> loads of data like the stuff in the attached file and then eventually
>>> couchdb would terminate.
>>>
>>>
>>> Respect,
>>>  Jason
>>>
>

Mime
View raw message