I have a Solaris 10 server running Subversion 1.5.1 served by Apache 2.2.9
sitting behind an f5 content switch. The content switch redirects all
traffic to the one Subversion server. In case of emergencies we can
manually redirect the content switch to a secondary server that is synced to
with svnsync. In general though, the traffic is all directed to one server.
This setup works fine in most cases. We have run into one odd problem
On Windows, if I checkout a working copy and try to store it on NAS via a
mapped drive, part way through the checkout I get the following error and
the checkout aborts.
svn: REPORT request failed on '/svn/REPONAME/!svn/vcc/default'
svn: REPORT of '/svn/REPONAME/!svn/vcc/default': Could not read response
body: An existing connection was forcibly closed by the remote host. (
The error does not occur when the working copy is stored locally, or when
the working copy is stored on NAS but I connect directly to the server
instead of the content switch.
A member of the network team here did a trace during checkout numerous times
and determined the following.
We noted in the traces that the client TCP Receive Window drops to 0,
indicating he can't receive any more data. He sends a zero window probe
about 8 times over about 45 seconds, after which time the server resets the
I cannot tell why the client TCP window goes to 0 when the traffic is
through the content switch, but not when it goes directly to the server. I
think a ticket should be opened with the vendor for the client program to
see if they've encountered this before.
Does anyone have any idea why this is happening? Should I change some
setting with Apache so the connection isn't closed? I realize NAS here is
slow, but we have a need for it. I just want to ensure that the checkout
isn't killed by the server.
Received on 2008-08-26 17:19:56 CEST