[distcc] Aborting jobs in status "Connect"

[Martin Pool]

On  1 Oct 2003, Christian Breimann <distccGroupEmail.20.cbreimann at spamgourmet.com> wrote:
> Hello,
> 
> I am using distcc to distribute jobs on several computers in an university 
> network. Some of them usually have problems like a very high load caused by 
> unsuccesful closed X-sessions or strange cups-daemon-processes. Due to 
> problems like this, some machines allow normal users like me to connect, 
> but not to log-in, i.e., resulting in "hanging" ssh- or rsh-sessions. Only 
> root is able to log-in and terminate such high-load-processes or reboot the 
> machine. I cannot ask my system administrator to do this several times a 
> day.
> 
> However, in these cases, distcc seems to have a similar problem. The 
> graphical monitor shows that distcc is in "Connect"-Status for several 
> seconds or even minutes without anything happening for that job. All other 
> machines get their jobs, finish them and get new jobs, only this one 
> machine hangs. After everything has been completed, I can terminate the 
> make-run using CTRL-C and start it again, so that the last jobs gets 
> finished this time on another machine.
> 
> So I wonder, whether distcc can do the following for me:
> If a distributed job remains in "Connect"-status for a certain amount of 
> time, perhaps a user-defined number of seconds or a default of 10 seconds, 
> distcc should kill this job, mark the machine as not available and 
> redistribute the job in the same way as if the machine is not reachable at 
> all.
> If "Send"-status is reached before this time limit everything should be 
> processed as before.

Hi,

Can you please turn on verbose client logging and show me at what
stage the client is hung?

-- 
Martin 
                               linux.conf.au -- Adelaide, January 2004