[torqueusers] Re: [Cluster Resources #4099] Torque resource manager does not send "command new" to Moab scheduler

Prakash Velayutham Prakash.Velayutham at cchmc.org
Mon Feb 16 13:36:27 MST 2009


Yup, may be that was it. Looks like even though Torque was trying to  
send a message to Moab to start a new scheduling cycle, Moab was  
ignoring it. I don't know if this event should have been logged by  
Torque though. Torque gurus, any comments about that?

Thanks,
Prakash

On Feb 16, 2009, at 3:24 PM, Douglas Wightman via RT wrote:

> I can't comment on the Torque side of things, but I do know that  
> there was a regression in Moab 5.3.1 that caused Moab to ignore any  
> events initiated by the resource manager.  I assumed this was the  
> problem on your system.
>
> - Douglas
>
> ----- "Prakash Velayuthum via RT" <moab- 
> support at clusterresources.com> wrote:
>
>> <URL: http://support.clusterresources.com/Ticket/Display.html?id=4099
>>>
>>
>> Hi Doug,
>>
>> I am surprised how this is a Moab issue. Why don't I see Torque
>> sending the "SCH_SCHEDULE_NEW" to Moab in Torque's logs, if the issue
>>
>> is Moab-related? Just curious.
>>
>> Prakash
>>
>> On Feb 16, 2009, at 2:16 PM, Douglas Wightman via RT wrote:
>>
>>> Prakash,
>>>
>>> This was a known regression in 5.3.1 and it has been fixed in
>>> snapshots since then.  You can download a snapshot or wait for 5.3.2
>>
>>> which will be released early this week.
>>>
>>> - Douglas
>>>
>>> ----- "Prakash Velayuthum via RT" <moab-
>>> support at clusterresources.com> wrote:
>>>
>>>> Ticket [Cluster Resources #4099] -- Ticket created by MoabCCH
>>>>      Queue: moab-support
>>>> Organization: Cincinnati Children's Hospital
>>>>    Subject: Torque resource manager does not send "command new"
>> to
>>>> Moab scheduler
>>>>      Owner: Nobody
>>>> Requestors: Prakash.Velayutham at cchmc.org
>>>>     Status: new
>>>> Ticket <URL:
>>>> http://support.clusterresources.com/Ticket/Display.html?id=4099 >
>>>>
>>>>
>>>> Hello,
>>>>
>>>> I am using Torque 2.3.6 with Moab 5.3.1 at my site. Here is the
>>>> strangest thing I see.
>>>>
>>>> When Torque RM is used with Torque scheduler, with "set server
>>>> scheduling = True", I see that the Torque server logs a
>>>> "PBS_Server;Svr;bmiclustersvcd1.cchmc.org;Scheduler sent command
>> new"
>>>>
>>>> message, but with Moab scheduler, everything works except a new
>> job
>>>> does not trigger this somehow. So Moab schedules the job every 30
>>>> seconds instead of with the new job trigger.
>>>>
>>>> Has anyone seen this? I have tested this like 10 times now and it
>> is
>>>>
>>>> consistent and reproducible. Every time I enable Torque scheduler,
>>
>>>> new
>>>>
>>>> jobs get scheduled immediately, but when Moab is enabled, Moab gets
>>
>>>> to
>>>>
>>>> new jobs only every 30 seconds.
>>>>
>>>> Thanks for any ideas.
>>>>
>>>> Prakash
>>>
>



More information about the torqueusers mailing list