[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [Condor-users] Swap space estimate reached! No more jobs can be run! Solution: get more swap space, or set RESERVED_SWAP = 0
- Date: Tue, 27 Sep 2005 12:58:30 -0500
- From: Erik Paulson <epaulson@xxxxxxxxxxx>
- Subject: Re: [Condor-users] Swap space estimate reached! No more jobs can be run! Solution: get more swap space, or set RESERVED_SWAP = 0
On Tue, Sep 27, 2005 at 05:32:57PM +0100, Chris Miles wrote:
> My ScheddLog is reporting
> 9/26 15:13:49 Sent ad to central manager for chris@xxxxxxxxxxx
> 9/26 15:13:49 Called reschedule_negotiator()
> 9/26 15:13:55 Activity on stashed negotiator socket
> 9/26 15:13:55 Negotiating for owner: chris@xxxxxxxxxxx
> 9/26 15:13:55 Checking consistency running and runnable jobs
> 9/26 15:13:55 Tables are consistent
> 9/26 15:13:55 Swap space estimate reached! No more jobs can be run!
> 9/26 15:13:55 Solution: get more swap space, or set RESERVED_SWAP = 0
> 9/26 15:13:55 0 jobs matched, 5 jobs idle
> RESERVED_SWAP has been set to 0 in the global condor_config file
> which has not made any difference. The daemons have all been reinstalled
> after that.
> I am using a Linux cluster with suse 9 installed. the cluster nodes contain no swap space.
> I can not find any information on this or any help on it.
> Any help is greatly appreciated.
Upgrade to 6.7. It works with the 2.6 kernel much better. Condor 6.6 can't
detect the memory or the swap space on a 2.6 kernel machine, and you can
only define MEMORY in the config file, not SWAP.
However, on linux/x86, that's not been a problem, because the bogus answer
we get back from the kernel is thankfully close enough, so it's just kind
of worked. Where we've seen problems with SWAP and the 2.6 kernel is on
other Linux platforms, like Linux/PPC. And then the problem is usually
that the Shadow won't start. If RESERVED_SWAP is set to 0, the
schedd shouldn't complain. Since you say that it's set in the global config
file, my guess is that it's being overrideden somewhere. Check with
condor_config_val and ask the schedd directly what it's set to:
cobalt(2)% condor_config_val -pool condor.cs.wisc.edu -schedd -name south.cs.wisc.edu reserved_swap -verbose
Defined in '/unsup/condor/etc/condor_config.global', line 82.
or set the SCHEDD_DEBUG to include D_FULLDEBUG and restart the schedd, it
will print out all of the values for swap space sizes it's using.
> Condor-users mailing list