Hello,
Lately we have been experiencing intermittent timeouts with our Subversion
operations. It does not happen initially, but after a while it starts
happening. Restarting Apache alleviates the problem, but it comes back after
a time. As you can imagine, this wreaks havoc.
Our operating environment:
- server - Windows 2003
- Apache 2.2.13
- Subversion server 1.6.3
- Subversion client 1.6.6
- mod_auth_sspi 1.0.4-2.0.58
- 200+ very active users, ~74K files
We have been doing a lot of things to try and mitigate the situation, but to
no avail. Changes have included:
- tweaking the memory module settings (WinNT MPM)
- packing the shards
- trimming hooks down to minimal activity
- monitoring system resources for spikes (none found, plenty of headroom,
no queueing, etc)
- examining the error and access logs (nothing interesting found)
One thing we cannot get much of a view into is the SSPI authentication
module (mod_auth_sspi). It does not seem to have any instrumentation. Has
anyone experienced timeouts or deadlocks with this module? Google isn't
turning up anything interesting. I've viewed the SVN interactions in
Wireshark. A normal sequence of operations is:
1. client: svn log request
2. server: 401, authorization required
3. client: send creds
In the hang scenario, we see just the initial client request (#1). Does this
help or hurt the theory that the mod_auth_sspi/AD interaction is causing the
problem? My next idea is to allow anonymous read access to the repo, which
may help prove that the authentication mechanism is someone responsible. If
nothing else, it should improve the performance.
Lastly, it might be worth mentioning that I have exposed the same SVN repo
on two different endpoints in Apache. That is, I have two location elements
(with different paths) but they both point to the same repo path. Is there
any problem with doing this?
Thanks for reading. Please let me know if you have any ideas.
Regards,
Dave Purrington
Received on 2010-01-11 20:25:43 CET