[svn.haxx.se] · SVN Dev · SVN Users · SVN Org · TSVN Dev · TSVN Users · Subclipse Dev · Subclipse Users · this month's index

Stalling repository via WEBDAV

From: Hannes Erven <h.e_at_gmx.at>
Date: 2005-05-26 12:46:35 CEST

Hi all,

I run a subversion gentoo ebuild (1.1.3, latest one marked "stable" by
gentoo) with a BDB repository.
Access is only allowed via Apache 2 / mod_dav_svn. There is a
post-commit hook script that uses the contributed perl script to send
out changes that accesses the repository directly on the file system.
As I felt this might be the source of the problem, I added a 60 seconds
delay to the script, so it waits a minute after the commit before
querying for changes.

Every now and then, the repository "stalls": requests to it (webdav and
direct) take ages to complete, e.g. a checkout that normally takes a
minute slowed down to 2 hours - with full cpu usage on the server, but
neither network traffic nor cpu usage on the client.
The server has a P4-HT cpu in it, and top shows 99% cpu load by an
apache process on one CPU, while the other one handles the other tasks.

My recovery procedure is as follows:
* shut down apache.
* kill / kill -9 any SVN leftover processes (most of the time, there
aren't any)
* svnadmin recover
* startup apache.

After that, the repository happily serves requests again at normal speed.

It seems, though I can't prove it, that this happens when there are
concurrent read and write accesses. Unfortunately, I cannot reproduce
the "stalling" repository.
svnadmin recover correctly tells me that the repo is locked, so I assume
that this is not a permission issue.

I appreciate any help on this - thank you in advance!

With best regards,

        -hannes

---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@subversion.tigris.org
For additional commands, e-mail: users-help@subversion.tigris.org
Received on Thu May 26 12:48:22 2005

This is an archived mail posted to the Subversion Users mailing list.