Bernard,
It sounds like a similar thing. When you restart the server all is
okay. The problem seems to be with how non-blocking I/O is handled by
pbs_server. The code is:
while ( (wrc = write(fd, buf, count)) < 0){
if (errno == EAGAIN) continue;
};
and when there is a real error it can never break out of the loop!
There do seem to be a few patches around to fix it (e.g. the OSC site)
but I'm concerned with making this change, as the OSC patch implies
there are a number of other things that should be done.
Frank
On Thu, 2003-12-11 at 11:49, Bernard Li wrote:
> Hi Frank:
>
> I have had issues when I bring UP nodes and the server just goes crazy :-)
>
> I suppose when your re-start the server it is okay?
>
> I think the problem is when a node goes down and PBS tries to schedule a
> job to the node, it just gets stuck...
>
> I have tested SGE and am quite happy with it, and I will in the near
> future be creating an OSCAR package for it. We should also be providing
> multiple batch systems support so that the users can choose to use.
>
> Cheers,
>
> Bernard
>
> Frank Crawford wrote:
>
> > Folks,
> > Has anyone else seen problems with pbs_server using 100% of a CPU when
> > one of the compute nodes goes down (or is taken offline)? I think this
> > is a problem with the cplantFR patch (as suggested by a couple of mail
> > items on the OpenPBS mailing list), but want to check with others first
> > before going down the wrong track.
> >
> > BTW, I've seen this in the OpenPBS RPM found in OSCAR 2.3, but given
> > the slow rate of change for OpenPBS, I don't expect it to be different
> > in others either.
> >
> > Thanks
> > Frank
> >
--
ac3
Suite G16, Bay 7, Locomotive Workshop Phone: 02 9209 4600
Australian Technology Park Fax: 02 9209 4611
Eveleigh NSW 1430
-------------------------------------------------------
This SF.net email is sponsored by: SF.net Giveback Program.
Does SourceForge.net help you be more productive? Does it
help you create better code? SHARE THE LOVE, and help us help
YOU! Click Here: http://sourceforge.net/donate/
_______________________________________________
Oscar-users mailing list
[EMAIL PROTECTED]
https://lists.sourceforge.net/lists/listinfo/oscar-users
