[GE users] Clearing errors: qmod -cq / qmod -cj issues

Reuti reuti at staff.uni-marburg.de
Wed Jan 25 20:17:29 GMT 2006


Hi,

Am 25.01.2006 um 10:32 schrieb Sebastian Stark:

>
> Am 24.01.2006 um 16:58 schrieb Reuti:
>
>>>
>>> Yes, I get the same error for queue instances that are not in  
>>> error state:
>>>
>>> neckar ~ % qmod -cq all.q at node143
>>> invalid queue "all.q at node143"
>>>
>>> And now I find that if I want to check if this queue instance has  
>>> an error it does not work at all:
>>>
>>> neckar ~ % qstat -f -q all.q at node143
>>> neckar ~ %
>>>
>>> This should return the status if this queue instance, right? This  
>>> works, however:
>>>
>>> neckar ~ % qstat -f | grep node143
>>> all.q at node143                  BP    2/2       14.97    lx24-amd64
>>> parallel.q at node143             BIP   13/16     14.97    lx24-amd64
>>>
>>> Something seriously broke in my sge installation since I upgraded  
>>> to 6.0u7, as it seems. It might also be a problem that I switched  
>>> from /etc/hosts to DNS based hostname resolution. Could it be  
>>> that SGE is confused about the fact that IP lookups return the  
>>> fqdn of the hosts now instead of just the host part?
>>>
>>
>> yes, there might be an issue, as the internal name known to SGE is  
>> only the short one. What is "qhost", "qconf -sel" saying? Maybe  
>> you have to add the machines now again (and you will see the FQDN  
>> also in qstat for each queue instance).
>
> qhost shows the hostnames only and qconf -sel shows the FQDN.
>
> If re-adding all hosts is the way to go I will try that as soon as  
> possible. Is there any way to have the domain name stripped from  
> the output of qstat et al? It's kind of ugly :/
>

I'm not aware of an option to suppress the FQDN, therefore I always  
echange the columns in /etc/hosts. With DNS of course I can imagine  
the long names of the queue-instances in the listing looking not  
really nice. If you use the "status" script from the file-exchange page

http://gridengine.sunsource.net/servlets/ProjectDocumentList

you will only see the queue name without the host. Specifying -m will  
give you the host (full name), but it would take only a small  
adjustment of the awk script to shorten it to reflect just the hostname.

-- Reuti

>
> -Sebastian
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: users-unsubscribe at gridengine.sunsource.net
> For additional commands, e-mail: users-help at gridengine.sunsource.net

---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe at gridengine.sunsource.net
For additional commands, e-mail: users-help at gridengine.sunsource.net




More information about the gridengine-users mailing list