[Systems] Pootle hung

[Bernie Innocenti]

On Tue, 2011-08-09 at 12:25 -0400, Chris Leonard wrote:

> There is a reasonable chance that I may have caused this issue.  We
> know that the pt-BR project (especially Honey) is messed up and can
> cause hangs.  I had successfully fixed an issue with the lang-da
> project by deleting the TamTam programs (while replacing with the new
> TamTamSuite  PO).  I took a chance and tried the same fix on pt_BR,
> but it appers that pt_BR is having some other problem.
> 
> I sincerely apologize for creating a problem and not being in a
> position to fix it myself..  I will work to get set up to log onto the
> Pootle server and be able to clean up my own messes, but in the
> interim, I will not try any more experiments with pt_BR.

Since Apache was spinning with 100% cpu, I suspect a bug in mod_python
(or are we using mod_wsgi now?), otherwise the errant process would have
been killed after some time.

Maybe some specific action can trigger this bug, or maybe every request
has a small chance of causing it. I bet it's the latter, because at the
FSF we've seen a similar race condition with gplv3.fsf.org and Zope.

Anyway, don't let such uncertainty hold you back! If you have shell
access to pootle.sugarlabs.org, you could restart Apache yourself in
case it hangs again.

We might also want to try upgrading mod_python or mod_wsgi, but this is
trickier because the machine is already a hybrid between two versions of
Ubuntu, for the purpose of satisfying Pootle's dependencies :-/

-- 
Bernie Innocenti
Sugar Labs Infrastructure Team
http://wiki.sugarlabs.org/go/Infrastructure_Team