kfogel@collab.net wrote:
> Someone in IRC had been seeing problems on the server side with a
> similar environment. Have you looked in your Apache httpd error logs
> to see if anything funny is going on? In particular I'm thinking of
> "exit signal", "illegal instruction", or anything like that. Below
> you quoted your access log, but not (I think) your error logs.
There were no entries in the {,ssl_}error_log or the global
/var/log/apache2/* files for that access. Just before I had
shut down apache and done a svnadmin recover and svnadmin verify.
After I started, I forgot to chown the files for httpd, so
there was an error for my first access, but then after the chown
it seemed happy, and the http2 went off doing its thing for
many minutes of run time in reponse to the svn update request. After
I sent my first email, I periodically watched, and the httpd
process kept on running even after the client gave up. It
looks like it did finish another 15 minutes later, so perhaps
a longer timeout would let it finish. However, there's no way
I could put up with such waits on a regular basis. cvs takes
only a minute or less for a similar operation.
> Also, have you tried using the 'svnserve' server, and avoiding Apache
> entirely? If you try that, does it behave differently?
>
No, I haven't tried that. If it isn't too troublesome, I'll
give it a try. Since my report, I've concentrated on getting
my cvs repository back up to date so I can go back to that for
this one project. (I have 5 other svn projects that all work
fine, but then they are not 3GB of data.)
> This is not expected behavior, no. 44010 files should not cause a 40
> minute propfind.
>
Actually it may have been a 40m REPORT, not PROPFIND, as in the
2nd log message. I'm not sure. Would a large number of files
(7588) in a single directory cause a problem? I am debating
whether to organize the files into subdirectories in the
cvs tree and then cvs2svn convert again. I was able to
svn update in that repository in other directories that were
not so large. Only that big one seems to take forever.
(Hence my query about non-linear algorithms.)
-Earl
---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@subversion.tigris.org
For additional commands, e-mail: users-help@subversion.tigris.org
Received on Sun Jan 23 20:50:45 2005