Severe svn log performance problems with apache

[mi...@agilent.com]

I know this issue has been discussed several times.  I am aware of the SVNPathAuthz Off feature that can be used.  But even with the SVNPathAuthz Off set, the performance is completely unacceptable compared to svnserve.

Here's my example running svn log

SVNPathAuthz On time - 1 min 53 s
SVNPathAuthz Off time - 42 s
svnserve - < 1s

I really want to be able to use apache authentication but this performance disparity is forcing me back to svnserve.  Any thoughts as to what I might change?  I'm using version 1.1.3 server running on redhat 8.0 box.  Here's my configuration.  I just comment out the SVNPathAuthz line out to test the enable case.


LoadModule dav_svn_module     modules/mod_dav_svn.so
 
# Only Needed if you decide to do "per-directory" access control.
LoadModule authz_svn_module   modules/mod_authz_svn.so
 
#
# Example location directive.
#
 
 
<Location /repos>
   DAV svn
   SVNParentPath /usr/repos
   SVNPathAuthz Off
   AuthzSVNAccessFile /repos/svn-access-file
 
   Satisfy Any
   Require valid-user
 
   AuthType Basic
   AuthName "Subversion Repository"
   AuthUserFile /repos/svn-auth-file
</Location>

Before anyone makes any suggestions as to modifications, I have have tried this simple version with no speed up.


<Location /repos>
   DAV svn
   SVNParentPath /usr/repos
   SVNPathAuthz Off
</Location>

Michael Rytting
Design Automation Engineer
Agilent Technologies
Colorado Springs Technology Center
1900 Garden of the Gods Rd.
Colorado Springs, CO 80907
work: 719-590-3708
fax:  719-590-3525

---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@subversion.tigris.org
For additional commands, e-mail: users-help@subversion.tigris.org

Re: Severe svn log performance problems with apache

[Ben Collins-Sussman <su...@collab.net>]

On Mar 24, 2005, at 12:27 PM, <mi...@agilent.com> wrote:

>
> SVNPathAuthz Off time - 42 s
> svnserve - < 1s

The discrepancy here is absurd;  I've never seen or heard of anything 
like it.  Maybe there's something unique going on here with your setup. 
  I can't even think of a reasonable explanation for what you're seeing.

How many revisions come down to the client when you run 'svn log'?

Have you done any network traces to see what's going on?  Any unusual 
observations with apache?



---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@subversion.tigris.org
For additional commands, e-mail: users-help@subversion.tigris.org