Hello everyone,

I have a computer cluster which has a CentOS operating system and uses SGE as a queue manager. The problem is that they have recently appeared when making a qstat -f an error of au ((a) larm, (u) nreachable), as shown below. So I understood that the problem is connection.

queuename qtype resv/used/tot. load_avg arch states
---------------------------------------------------------------------------------
all.q@compute-0-0.local BIP 0/0/20 -NA- linux-x64 au
---------------------------------------------------------------------------------
all.q@compute-0-1.local BIP 0/0/20 -NA- linux-x64 au
---------------------------------------------------------------------------------
all.q@compute-0-2.local BIP 0/0/20 -NA- linux-x64 au


I have tried to restart the cluster and update the operating system, but it has not corrected the problem. I have also tried to restart the sge through the command ./sgemaster start (which tells me that it is on) and ./sgeexecd start (which tells me that it is starting it). Despite all this, the error persists . Can you think of how I could solve my problem?

Thank you very much to all.

Regards

Rafael

What monitoring do you have for network, disk or memory problems? A cluster like that should have Nagios or Munin or something.

Does the sge_execd run on the compute nodes?

Hi everybody,

I execute "pgrep -l sge" and exit is "6074 sge_qmaster 6271 sge_execd" I think that sge_execd are running on the compute nodes. Respect to the problems, when I execute qstat -j command I obtain:

scheduling info: queue instance "all.q@node_name_0" dropped because it is temporarily not available
queue instance "all.q@node_name_1" dropped because it is temporarily not available
queue instance "all.q@node_name_2" dropped because it is temporarily not available
All queues dropped because of overload or full


Thanks so much.

Regards

Hello everyone,

I finally found the fault. It was that the sge_exced were not running on the calculation nodes, so you only had to access node to node and restart it with the command ./sge_exced.
In this way the "au" fault is removed from the queue manager.


Thank you very much to all.
Regards.
RP