Hi Dave,

Thanks for your reply.  I have filed JIRA: 
https://issues.apache.org/jira/browse/COUCHDB-1856 to report this problem.

And one question by the way: I guess the problem is caused by the time-out of 
the http-connection. Is there anyway I can
configure the replication time-out for one document. For example, I'd like the 
http-request of one doc stay without time-out even
for more than 10 minutes. 


-----Original Message-----
From: Dave Cottlehuber [mailto:[email protected]] 
Sent: Thursday, July 18, 2013 11:42 PM
To: [email protected]
Subject: Re: What's the best way to replicate a couple of big databases

On 18 July 2013 15:59, Rao, Zhiqing <[email protected]> wrote:
> Hi Guys,
>
> I have a dozen of big databases which are about 100 Mb ~ 2GB, in win7 
> host( I can move to linux server later) , and want them to be replicated 
> daily to another Couch server. However, I tried the following ways:
>
> 1)      Create the corresponding docs in the _replicator database in futon, 
> it works one by one, but sometimes the erl.exe crash and I
>
> Have to restart couchdb;

Welcome Zhiqing,

This shouldn't happen if you're running couchdb as a service, and these 
databases are definitely smaller than ones I've used in the past without issues.

- are you running as a service?
- exactly which windows is this on?
- what couchdb release & what erlang release?
- can you switch debug level logging on (for a brief period of time, it doesn't 
require a restart but it does generate a lot of logs)

curl -XGET http://admin:passwd@localhost:5984/_config/log/level
- note this parameter down and use later on as your_level below curl -XPUT 
http://admin:passwd@localhost:5984/_config/log/level -d '"debug"'
- initiate your replications
- after issue turns up, apply your old config back again curl -XPUT 
http://admin:passwd@localhost:5984/_config/log/level -d '"your_level"'

NB I'm not at a windows box right now to confirm if the quoting works or not, 
check http://wiki.apache.org/couchdb/Quirks_on_Windows for the details.

Then please open a JIRA ticket with the logs; you should confirm there are no 
passwords or related confidential info (network names etc) present in the logs, 
along with curl -XGET http://admin:passwd@localhost:5984/_config/replicator and 
the replication you are using, and a rough estimate of average size of docs & 
details on attachments, and any other useful info like %mem used etc.

A+
Dave

Reply via email to