[HCoop-Discuss] When is response time on site going to get any better?

[Davor Ocelic]

On Sun, 11 Oct 2009 17:46:13 -0400
Clinton Ebadi <clinton at unknownlamer.org> wrote:
> 
> docelic performed the changes earlier today and apache has locked up
> since then.

That was supposed to read "hasn't", but yeah ;)

> It turns out it wasn't the afs client causing the problem
> (although the afs client changes have reduced load on the fileserver
> slightly so they were not made in vain), but rather a bad interaction
> between a broken cgi and apache.

Right. As people following -sysadmin might already know, one of our
members had a broken Wiki installation which was not working at all,
except for the unfortunate side-effect of choking our web server more
and more every time a request came to it.

Various automated web spiders that crawl the web were making enough
requests to that website, that our Apache instance would come to a
complete halt and need a restart every 20 or 30 minutes.

> We've also moved spam processing onto hopper significantly reducing
> the load on deleuze. This has made IMAP at least noticeably faster,
> and the updated spamassassin rules seem to have reduced false
> negatives as a nice side effect.

Yes. We've also done a bunch of other improvements while hunting
for the Apache bug for everyone's benefit.

> Services should all be stable and fast now, but if missed something
> let us know.

We're back on track with no outstanding issues.

What's important as well is that thanks to the work of Clinton Ebadi,
Frank Bynum, Richard Darst, Lorenzo Villani and others (apologies if
I missed anyone's name), we were able to re-establish a situation where
a significant amount of time is being dedicated to HCoop's technical
and organization management, and where we are progressing forward on
multiple fronts.

Enjoy everyone,
Thanks,
-doc