On Thu, 28 Dec 2006, Daniele Cesini wrote:
> Hi, I've got the following problem with an lcg lsf ce:
> globus-job-run <ce_host> /usr/bin/whoami works fine so authentication on
> the ce is not a problem
> but
> globus-job-run <ce_host>:2119/jobmanager-lcglsf -q <queue> /usr/bin/whoami
> or
> globus-job-run <ce_host>/jobmanager-lcglsf -q <queue> /usr/bin/whoami
> returns nothing, it does not hang, but no answer is returned.
Did you spell the queue name correctly?
If the queue is unknown, you will get back an empty response...
That will be another Wiki entry soon.
> On the gatekeeper log there are no errors. The command whoami is not
> submitted to lsf, but local lsf submission with bsub from the user
> account works fine.
> WN are set up correctly because the same WNs accessed by the gLite CE do
> not give any problem.
>
> The only strange lines that I see on the gram_job_mgr_16885.log file in
> the user home directory are like these, but not very meaningful to me:
>
> 12/28 17:51:49 JMI: poll_fast: Monitoring file looks out of date.
> Reverting to normal polling
> 12/28 17:51:49 JMI: poll_fast: returning -1 = GLOBUS_FAILURE (try Perl
> scripts)
> Thu Dec 28 17:51:19 2006 JM_SCRIPT: Cache too old for this job
> (unknown): make_a_poll_query() returning 0
Those messages are normal.
> I do not what else I can check.
> Has someone any ideas?
As the local account to which you are mapped, run this on the CE:
env | sort > /tmp/env-1
Then compare with the result of this:
globus-job-run <ce_host> /usr/bin/env | sort > /tmp/env-2
Let me know the hostname and I can have a look.
