Re: [Condor-users] COD troubleshooting

[Renzo Borgatti <borgatti@xxxxxxxx>]

On Dec 27, 2006, at 3:43 PM, Dan Bradley wrote:
> Unfortunately, COD doesn't currently support transferring any files,
> including the x509 proxy file.  Is it possible for you to rely upon a
> shared filesystem for this purpose?

No unfortunately, because the Condor installation is glide-in based  
(most probably the strongest use-case to have gLExec in place). And I  
don't know how the starter search for the X509. Is this done using  
the shadow on the head node? Or should the x509 be somewhere on the  
worker?

> I haven't thought through what permissions would be necessary in  
> order for this to work for gLExec.

Afaik, a COD is just a special kind of job with short latency because  
there is no negotation and preemption is always on. Probably it  
should work as any other jobs when gLExec is active, so the startd is  
doing the right thing. Without knowing the internals of the  
mechanism, I can't say if it just needs to have an env variable in  
place to work or something else.

Can you or someone else confirm that the COD feature is not yet ready  
to be used in conjuction with gLExec? I just want to be sure that  
there is nothing else I can do from my side.

Renzo

> --Dan
>
> Renzo Borgatti wrote:
> > Thanks Dan. I added both 'IWD' and 'User' without success. But I
> > found a startd not crashing and a StartdLog more verbose:
> >
> > 12/27 14:33:52 DaemonCore: Command received via TCP from
> > condor@fcdfcaf445 from host <131.225.240.106:35843>
> > 12/27 14:33:52 DaemonCore: received command 1000 (CA_AUTH_CMD),
> > calling handler (command_classad_handler)
> > 12/27 14:33:52 Serving request for CA_ACTIVATE_CLAIM by user 'condor'
> > 12/27 14:33:52 vm2: State change: Suspending because a COD job is now
> > running
> > 12/27 14:33:52 vm2: Changing activity: Retiring -> Suspended
> > 12/27 14:33:52 vm2: cannot use glexec to spawn starter: no proxy (is
> > GLEXEC_STARTER set in the shadow?)
> > 12/27 14:33:52 vm2: writeJobAd: Write_Pipe failed
> > 12/27 14:33:52 vm2: ERROR: exec_starter returned 0
> >
> > Looks like gLExec activation is used also to activate COD. I didn't
> > mention before that gLExec is active in my configuration. The error
> > has something to do with the X509 proxy not present. Is the mechanism
> > to transport X509 the same as universe=grid jobs? Is it possible to
> > specify with what X509 proxy the COD should run under?
> >
> > Thanks
> > Renzo
> >
> >
> > On Dec 27, 2006, at 12:29 PM, Dan Bradley wrote:
> >
> > > Hello,
> > >
> > > I have a hunch that some of the ClassAd attributes that the COD  
> > > manual
> > > claims are optional are actually required.
> > >
> > > --Dan
> > >
> > > Renzo Borgatti wrote:
> > >
> > > > Hi,
> > > >
> > > > I have a problem activating claims using COD (Condor 6.9.0).  
> > > > This is
> > > > what I'm doing:
> > > >
> > > >
> > > > > condor_cod request -addr "<131.225.212.148:39446>" -classad ci.out
> > > > Successfully sent CA_REQUEST_CLAIM to startd at
> > > > <131.225.212.148:39446>
> > > > Result ClassAd written to ci.out
> > > > ID of new claim is: "<131.225.212.148:39446>#1167216341#4"
> > > >
> > > >
> > > > > condor_cod activate -id "<131.225.212.148:39446>#1167216341#4" -
> > > > classad ci.out -jobad TestCod
> > > > Attempt to send CA_ACTIVATE_CLAIM to startd <131.225.212.148:39446>
> > > > failed
> > > > Reply ClassAd returned 'Failure' but does not have the ErrorString
> > > > attribute
> > > >
> > > > On the worker node, I can see the following two lines in the
> > > > StartdLog right before crashing:
> > > >
> > > > 12/27 11:50:05 DaemonCore: Command received via TCP from
> > > > condor@fcdfcaf444 from host <131.225.240.106:45123>
> > > > 12/27 11:50:05 DaemonCore: received command 1000 (CA_AUTH_CMD),
> > > > calling handler (command_classad_handler)
> > > >
> > > > while in the MasterLog:
> > > >
> > > > 12/27 11:55:30 The STARTD (pid 15721) died due to signal 11
> > > > 12/27 11:55:30 All daemons are gone.  Exiting.
> > > > 12/27 11:55:32 **** condor_master (condor_MASTER) EXITING WITH
> > > > STATUS 0
> > > >
> > > > TestCod is a file with the following 2 lines:
> > > >
> > > > Cmd="/bin/ps"
> > > > Args="-aux"
> > > >
> > > > Am I using condor_cod the right way? Is there a way to have more
> > > > debugging information to understand what happened?
> > > >
> > > > Thanks
> > > > Renzo
> > > > _______________________________________________
> > > > Condor-users mailing list
> > > > To unsubscribe, send a message to condor-users-request@xxxxxxxxxxx
> > > > with a
> > > > subject: Unsubscribe
> > > > You can also unsubscribe by visiting
> > > > https://lists.cs.wisc.edu/mailman/listinfo/condor-users
> > > >
> > > > The archives can be found at either
> > > > https://lists.cs.wisc.edu/archive/condor-users/
> > > > http://www.opencondor.org/spaces/viewmailarchive.action?key=CONDOR