Hi Jaap, thanks. We did commit a recent change to the replicator that
would make high-latency pull replications take significantly longer
than before. Specifically, we stopped requesting multiple documents
in parallel. This should only be a temporary situation; we wanted to
make sure we had the replicator memory usage totally under control
before turning the parallelization back on.
I'm not always sure why the remote CouchDB server sometimes chooses to
close the connection during replication. You're not the first person
to see this happen, and in other cases these connection_closed errors
have occurred when replicating inside a LAN.
Did you know you can check the status of the replication in Futon? If
you go to the "Status" page on the server where you initiated the
replication you should see a replication task that updates every time
it processes 100 docs on the source. Best,
Adam
On Mar 20, 2009, at 8:59 AM, Jaap van der Plas wrote:
> Thanks, problem seems to be the connection:
>
> [Fri, 20 Mar 2009 12:45:47 GMT] [info] [<0.262.0>] retrying couch_rep
> HTTP get request due to {error, connection_closed}:
> http://localhost:5985/db/303ccbc3394aad84ff4bbce74470036b?revs=true&latest=true&open_revs=
> ["4-3427878696"]
>
> [Fri, 20 Mar 2009 12:47:10 GMT] [info] [<0.262.0>] recording a
> checkpoint at source update_seq 653
>
> [Fri, 20 Mar 2009 12:47:11 GMT] [info] [<0.262.0>] retrying couch_rep
> HTTP put request due to {error, connection_closed}:
> http://localhost:5985/db/_local%2Ff7a03c6321dd9aa5652120b88cd4fb4d
>
> [Fri, 20 Mar 2009 12:47:12 GMT] [info] [<0.68.0>] 127.0.0.1 - - 'POST'
> /_replicate 200
>
> After this it did replicate correctly (for the first time.)
> Replication was to an empty database btw. Not sure why this problem
> never happened in previous versions though... Could it be because the
> connection is relatively slow? (Remote server is in the US, I'm from
> the Netherlands. Loading the remote database url in a browser takes
> about 1 second.)
>
>
> On Fri, Mar 20, 2009 at 12:37, Adam Kocoloski <kocolosk@apache.org>
> wrote:
>> On Mar 20, 2009, at 6:07 AM, Jaap van der Plas wrote:
>>
>>> After upgrading to the new database format (using a script that
>>> simply
>>> read docs from the old database, and created them as new documents
>>> in
>>> the new database) I am no longer able to replicate a certain
>>> database
>>> from a remote server to my local machine. Using an ssh tunnel I
>>> access
>>> the remote database on localhost:5985.
>>>
>>> Using the replicator interface on the local instance of couchdb, I
>>> am
>>> able to replicate from my local database, to the remote database.
>>> However, the reverse doesn't work: the _replicate post request
>>> takes a
>>> very time / doesn't seem to finish at all (seen with firebug.) The
>>> database in question is 1mb in size (compacted.) If I try
>>> replicating
>>> with a db with 1 doc in it it does work both ways. Where should I be
>>> looking for the cause of this problem? Is it possible there is some
>>> offending/corrupt document in the database?
>>>
>>> regards,
>>> Jaap
>>
>> Hi Jaap, is there anything in the logfile for the local instance?
>> The
>> replicator is usually pretty chatty about any problems. Best,
>>
>> Adam
>>
|