Ben Collins-Sussman <sussman@collab.net> writes:
> Garrett Rooney <rooneg@electricjellyfish.net> writes:
>
> > > svn_error: #21081 : <RA layer request failed>
> > > neon: Could not read status line: connection timed out.
> >
> > Looks like the damn repos hung again... db_recover will need to be
> > run before anyone can do anything. We need to find this bug...
>
> This is getting worse. It's happening approximately every two days
> now.
>
> I logged into svn.collab.net as user 'svn', and saw 45+ httpd
> processes. Don't know if this is normal or not, but 'apachectl stop'
> tried to stop a non-existent PID to no avail. I had to 'killall
> httpd', which killed about 40 of the processes. The other 5 could
> only be killed as user root.
>
> Then I ran 'db_recover -ve' on the repository. According to the
> 'time' command, it took 16 minutes. This is very disturbing as well,
> since this bug used to happen about once per month, and the recovery
> process took only about 3 seconds. Whatever the bug is, it's not
> scaling well anymore.
Benny, this happened to me the other day as well ('cept I didn't wait
the full 16 minutes...I restored from backup). Just letting you know
that this has now happened twice -- that tonight's occurance wasn't
just a one-time freak accident.
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Sat Oct 21 14:36:58 2006