On Fri, May 09, 2003 at 12:24:15AM +0200, Wichert Akkerman wrote:
> Previously Thom May wrote:
> > I'm sure Wichert will be able to shed more light on the locking
> > problems he's come across
>
> The behaviour we see is this: when we have both read-only access
> through the apache2 svn module and write-access through svnserve it
> frequently happens that the apache module gets stuck and does not
> release a lock (lsof shows an open file handle on locks/db.lock). This
> means that the a commit using svnserve will block indefinitely, until
> I kill off apache which releases the lock. Occasionaly however the
> repository is left in a bad state during that process and I have to
> recover it.
>
> This problem is very noticable with our linux kernel repository so it
> seems that the size or complexity of the repository does play a role
> in the likelyhood this problem is hit.
I tried clicking on the linux kernel repository link when I saw the
announcement and I think I got a 404 error or something and then tried
the viewcvs link. The viewcvs link gave a Python backtrace mentioning
an out of memory error as far as I remember.
Couldn't think just be a case of not enough BDB locks in
repos/db/DB_CONFIG? I seem to remember that causing out of memory
errors. Maybe some other values in DB_CONFIG need to be tweaked.
--
Michael Wood <mwood@its.uct.ac.za>
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Fri May 9 16:26:00 2003