couchdb-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Chris Anderson" <>
Subject Re: doc_count set to zero even though documents remain
Date Tue, 22 Jul 2008 20:22:39 GMT
I've seen the same thing -- the easiest way I've had to replicate it
is to rename the database.couch file, after having compacted, and then
compact again. I thought that was a weird enough case that it didn't
matter, but it looks like Peter has been able to cause it without such
odd usage.

On Tue, Jul 22, 2008 at 8:30 AM, Peter Eddy <> wrote:
> Thanks,
> It's definitely not a browser cache issue, I first noticed it testing
> my Lisp -> CouchDb library, I then verified the condition using a
> browser that had never seen the CouchDb server machine before.
> All my tests are automated so I'll just re-run them and see if I can
> reproduce the issue.
> - Peter
> On Mon, Jul 21, 2008 at 9:23 PM, Damien Katz <> wrote:
>> I've not heard of other reports of this. To make sure it's not a browser
>> cache issue, check the count from another browser by pointing the browser at
>> the db url and see what is returned.
>> -Damien
>> On Jul 21, 2008, at 6:32 PM, Peter Eddy wrote:
>>> Hi,
>>> I ran into this a few days ago and haven't had a chance to look into
>>> it, but I thought I'd mention it any way.
>>> I was doing some performance testing on some new hardware, I created 1
>>> million test documents (exactly) and then ran various multi-threaded
>>> tests to time document fetch time. I then compacted my database and
>>> shutdown CouchDb well after compaction had finished.
>>> When I restarted CouchDb I found that even though the 1 million
>>> documents are still there, the doc_count value is zero. I can still
>>> fetch the documents without any problem.
>>> I'm using CouchDb 0.8-incubating and Erlang R12B-3 on a 4 core AMD64
>>> Linux box with kernel 2.6.24 SMP. I'm not using replication. I
>>> encountered no errors while loading the database with the test
>>> documents or while running the tests.
>>> thanks,
>>> Peter

Chris Anderson

View raw message