[Pound Mailing List] Re: PUT requests triggering "error copy chunk cont: Broken pipe"

Sorry, I forgot to mention, pound also eventually logs this when it
kills the connection:

2014-01-17T16:21:29+00:00 localhost pound: [ID 702911 daemon.notice]
(82) e500 for 72.2.112.156 response error read from
127.0.0.1:15984/PUT /registry/blade?new_edits=false HTTP/1.1:
Connection timed out (300.720 secs)

But that seems really weird, since CouchDB received the request, and
sent a response already.


On Fri, Jan 17, 2014 at 8:24 AM, Isaac Schlueter <[email protected]> wrote:
> I have pound running in front of a CouchDB server on the same machine.
>
> When I try to do a PUT of a ~4MB document, pound eventually kills the
> connection without sending a response, logging this message:
>
> 2014-01-17T16:08:49+00:00 localhost pound: [ID 702911 daemon.notice]
> (7e) error copy chunk cont: Broken pipe
>
> I can see in the CouchDB logs that the PUT made it to the backend
> server, and that it responded with a 201 response code:
>
> [info] [<0.118.0>] 72.2.112.156 - - PUT /registry/blade?new_edits=false 201
>
> The code that's doing the PUT logs this:
>
> PUT /registry/blade?new_edits=false
> { 'content-length': 4739164,
>   'content-type': 'application/json' }
>
> I've also tried sending a connection:close header, but that makes no difference.
>
> Previous searches said that this could be due to low timeout values.
> I have this in my pound.cfg file:
>
> Alive           300
> Client          300
> TimeOut         300
> ConnTO          300
>
> It takes CouchDB about 10 seconds to respond to the request, so a 300s
> timeout should be more than enough.  However, even after CouchDB sends
> the 201 response, pound waits for 5 minutes, and then kills the
> connection, logging the message above.
>
> Is there something else I can do to try to debug this?
>
> Thanks.

--
To unsubscribe send an email with subject unsubscribe to [email protected]
Please contact [email protected] for questions.