Hi Lucas,
You identified the problem correctly: network connection drop. All the
errors are coming from that problem. Just to explain a bit, once you
modify a document, your "push" replication tries to push the new
document version in the targeted database, but due to the no connection
condition, dies after a number of retries. I would suggest to use simple
replication and not continuous one if your connection is unstable. That
is, every given period, check if the connection is up and send a
replication request if the network allows it.
Another option is to check your connection periodically (you can
automatize that by the mean of a simple ping for example) and if the
connection is dropped, stop the continuous replication (or let it die)
and start it again when the network conditions allow that. But I would
put an watchdog on the network connection and not periodically restart
of the replication.
As far as I could notice from the log, no specific document creates this
problem, but only the drop in network connection. Therefore, I wouldn't
worry about the documents correctness for the time being.
CGS
On 12/26/2011 02:18 AM, Lucas Chiesa wrote:
Hello,
We are using couchdb for a project and so far it has been working
great. However we have stumble uppon a problem we cannot seem to fix.
I'll try to describe the problem and I'm attaching the logs, without
any "filtering" because I don't understand what's important and what's
not.
We have 3 dbs in one way continuous sync and it all worked OK until,
for some reason some documents seen to stop the synchronization of one
of them.
If we manually sync that document like this:
$ curl -H 'Content-Type: application/json' -X POST
http://localhost:5984/_replicate -d ' {"source": "machines", "target":
"http://10.8.0.1:5984/machines";,
"doc_ids""f9dfe9e74c9818c136bd5378e40012fa"]} '
{"ok":true,"start_time":"Mon, 26 Dec 2011 00:02:09
GMT","end_time":"Mon, 26 Dec 2011 00:02:10
GMT","docs_read":1,"docs_written":1,"doc_write_failures":0}
It works fine, and then the continuous sync works, until we modify
this document (for example saving it using futon, creating a new rev).
Please note that we were issuing the continuous sync command once per
minute because of a bug in a cron job. It was supposed to be once per
hour, and that's because we saw that if the computer lost internet for
a long time, couchdb dropped the sync.
Most certainly I have not explained the problem correctly, but I don't
understand it correctly, I hope the logs help better.
Warning: The logs (uncompressed) are almost 40M big....
Thank you very much in advance,
Lucas
"Oh no, not again."
- A bowl of petunias on its way to certain death.
