[OGo-Users] Webui killed by watchdog?

[Helge Hess]

On 12.02.2008, at 16:02, Albrecht Dre=DF wrote:
> Well, I have a script, called by cron, which reads using wget the =20
> login page, and if this fails, calls the init script with restart.

OK, so this is what triggers the TERM signal.

> It sounds as if this is not the right approach, as the watchdog is =20
> supposed to restart the webui automagically.  Is that true?

Auto-Restart on crashes, yes.

>  How does it detect any "hangs"?


No, it can't detect hangs. I think snsd can, but won't restart a =20
process either.


So your script is quite fine. So back to your original questions:
> Questions:
> - what is the reason for the watchdog timeout?

There is no timeout, the TERM signal is triggered by your script.

> I /might/ be related by OgO trying to contact an *extremely* slow M$ =20=

> Exchange server which sometimes needs 2 minutes (!!) to return a =20
> message list (this software is complete crap!).  Any way to increase =20=

> this timeout?

I think the socket timeouts are hardcoded, you might be able to change =20=

them in the sources and recompile.
BUT: this only makes sense if you run multiple OGo processes anyways. =20=

Otherwise your slow server will block all other sessions.

> - if the watchdog kills a process, is it possible to restart it =20
> immediately?


The watchdog only kills a process when its requested to do so.


Hm, maybe your script does an 'ogo stop' instead of 'ogo restart'? Or =20=

do you use kill in your script? In this case sending HUP might make =20
Hector restart the child process.

Greets,
   Helge
--=20
Helge Hess
http://www.helgehess.eu/=