I have done a little bit of investigation -- here's the situation as
relayed to us by David Gibbons at Planet Argon:
> Within the last few weeks it was up, we were seeing cpu usage
> and memory usage skyrocket, IT'd peg the cpu at 100%, peg memory
> usage at 2 gigs. Disabling it was the only way we could keep
> the server usable for the other people on it. This happens with
> trac via mod_python, cgi, and tracd.
... and in a later message he said ...
> ...it seems programmatic in nature as it takes a very short period
> of time for it to start blowing out our resources. I didn't
> find anything conclusive in my testing, except that it occurs
> under cgi, mod_python as well as tracd.
Just to complete the picture, David said this is what our installation
looks like:
Trac .10 from subversion
Sqlite 3
Akismet for spam protection
This weekend I sent the following back to David, Scott, and Kevin:
> I was digging through Trac's mail list, and I found references to
> two tickets that might be similar to our problem:
>
>
http://trac.edgewall.org/ticket/3779 >
http://trac.edgewall.org/changeset/4127 >
> So if one of our attacking spambots posted changes to a page
> that triggered either of these bugs, we would see the CPU
> spike whenever someone viewed that page. None of the messages
> on the list mention excessive memory usage, so perhaps these
> bugs aren't the same as what we are seeing.
>
> These changes have been rolled into branches/0.10-stable of Trac.
So we basically need someone with both time and a shell account on the
machine to test fixes for our Trac instance. Sprewell and I have some
time, but no shell account.
Anyone with experience fixing broken Trac installations, now would be a
great time to offer up any advice gained from the experience. :-)
Best regards,
Tim
Tim Connor wrote:
> So what's the status on Trac looking like? It's just starting to get
> a wee bit annoying to be using Typo with its site down indefinitely
> when there are things on there that people might need that have even
> been linked to by the devs, but are unavailable - say like the fix for
> the expired API key? That's just one example, though.
>
> Thanks,
> Tim
> _______________________________________________
> Typo-list mailing list
> Typo-list at rubyforge.org
> http://rubyforge.org/mailman/listinfo/typo-list
--
Timothy Freund
http://digital-achievement.com
http://tim.freunds.net/thoughts