[Tilesathome] Weird failure of T at H client

[Christopher Schmidt]

On Thu, Nov 15, 2007 at 11:11:57PM +0800, D Tucny wrote:
> On 15/11/2007, Christopher Schmidt <crschmidt at metacarta.com> wrote:
> >
> > On Thu, Nov 15, 2007 at 11:58:10AM +0000, Robert (Jamie) Munro wrote:
> > > -----BEGIN PGP SIGNED MESSAGE-----
> > > Hash: SHA1
> > >
> > > Christopher Schmidt wrote:
> > > > On Thu, Nov 15, 2007 at 12:52:21AM +0000, Robert (Jamie) Munro wrote:
> > > >> -----BEGIN PGP SIGNED MESSAGE-----
> > > >> Hash: SHA1
> > > >>
> > > >> I just looked at my T at H client and found:
> > > >>
> > > >> I've restarted it (by typing ./tilesgen.pl loop), and it seems fine
> > now.
> > > >>
> > > >> Any ideas what the cause was? Should I worry?
> > > >
> > > > The server was down for about two hours earlier today; this was the
> > > > cause.
> > >
> > > Wouldn't that mean that most T at H clients would have stopped? Looking at
> > > the Munin graphs, there is a 2 hour gap between 1700 and 1900, but after
> > > 1900 T at H is quickly back up to full speed.
> >
> > Maybe I missed something in your email. The t at h clients, as far as I
> > know, all died as soon as the server did. Are you saying that yours did
> > not die during the server downtime ,but died at some time later?
> 
> 
> I think his point is that when the server was down, his client died, but,
> that looking at the munin graphs, it appears that most other people's seemed
> to just continue after the return of the server...
> 
> I'm guessing that just deelkar restarting his renderers is enough to make
> take t at h to as fast as the API will allow it, making the graphs look like
> they do...

Well, I restarted 8 renderers immediately after the server downtime, and
I know that several other people did the same -- the t at h queue showed
many clients coming back relatively quickly, but I don't think anyone
did it automatically in any way.

So, I think enough renderers are just observant that we picked up
quickly :)

Regards,
-- 
Christopher Schmidt
MetaCarta