[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
RE: [Condor-users] Requeueing a job from an execution client?
- Date: Tue, 29 Jun 2004 14:47:50 -0700
- From: "Ian Chesal" <ICHESAL@xxxxxxxxxx>
- Subject: RE: [Condor-users] Requeueing a job from an execution client?
So I just noticed that the user who owns files on disks for a job is
'condor-reuse-vm1' -- can someone on the condor team comment on how safe
it is to rely on the user name of files created on disk like this to
determine the virtual machine a job is executing on?
> -----Original Message-----
> From: condor-users-bounces@xxxxxxxxxxx
> [mailto:condor-users-bounces@xxxxxxxxxxx] On Behalf Of Ian Chesal
> Sent: June 29, 2004 3:14 PM
> To: Condor-Users Mail List
> Subject: RE: [Condor-users] Requeueing a job from an execution client?
> The -help from condor_vacate says:
> "If you specify a virtual machine(for example,
> "vm1@hostname"), only that machine will be vacated."
> However there isn't a CONDOR_VM environment variable present
> when a vanilla job runs. I dumped the environment from my
> perl script and I'm
> Environment is:
> BATCHSYSTEMFTPSERVER = 192.168.0.55
> COMPUTERNAME = TTC-TESTBATCH01
> COMSPEC = C:\WINNT\system32\cmd.exe
> CONDOR_INHERIT = 1108 <192.168.0.100:2820> 0 0
> NUMBER_OF_PROCESSORS = 2
> OS = Windows_NT
> PATH =
> PATHEXT = .COM;.EXE;.BAT;.CMD;.VBS;.JS;.WS
> PROCESSOR_ARCHITECTURE = x86
> PROCESSOR_IDENTIFIER = x86 Family 6 Model 5 Stepping
> 2, GenuineIntel
> PROCESSOR_LEVEL = 6
> PROCESSOR_REVISION = 0502
> PROGRAMFILES = C:\Program Files
> PROMPT = $P$G
> PWD = C:/condor/execute/dir_1108
> SYSTEMDRIVE = C:
> SYSTEMROOT = C:\WINNT
> TEMP = C:\WINNT\TEMP
> WINDIR = C:\WINNT
> _CONDOR_SCRATCH_DIR = C:\condor\execute\dir_1108
> I did play with the exit status stuff but it wasn't playing
> well with my perl script. It seemed to end the job no matter
> what exit status I returned.
> > -----Original Message-----
> > From: condor-users-bounces@xxxxxxxxxxx
> > [mailto:condor-users-bounces@xxxxxxxxxxx] On Behalf Of Erik Paulson
> > Sent: June 29, 2004 2:59 PM
> > To: Condor-Users Mail List
> > Subject: Re: [Condor-users] Requeueing a job from an
> execution client?
> > On Tue, Jun 29, 2004 at 11:44:47AM -0700, Ian Chesal wrote:
> > > Is it possible to requeue a job, send it back to the
> > scheduler, while
> > > it's executing on the client from that client, from within
> > the job. If
> > > you had a control script that represented your job and
> you hit some
> > > bad state could you send the job back to the scheduler?
> > Have the job exit, and set your exit requirements
> > appropriately. See the on_exit_remove option to condor_submit:
> > I'm thinking
> > condor_vacate is what I'd want to call but I can't see how
> a job would
> > know which VM on a machine it was running on...
> Conveniently, condor_vacate doesn't know about VMs anyway :)
> (There's also the CONDOR_VM environment variable that the job
> should have in it's environment, see
Condor-users mailing list
Condor-users mailing list