ID #1032

How do I integrate externally executed UGE/SGE server with Bright Cluster Manager 6.0?

  • Mount the $SGE_ROOT directory from your external server that is to be managed by the Bright Cluster Manager (BCM) head node. Make sure that q* commands (qstat, qconf, ...) on the head node can be called successfully. 
  • Add the following parameters to the AdvancedConfig section of /cm/local/apps/cmd/etc/cmd.conf:
    1. SGERoot=/path/to/your/mounted/sge/
    2. SGEVar=/path/to/your/mounted/sge/
    3. AlwaysAssumeSGERunning=1 
  • Restart the cmd daemon on the head node
  • Assign the sgeserver role to the head node in cmsh or cmgui. For example, using cmsh:

[root@bright60 ~]# cmsh
[bright60]% device
[bright60->device]% use master
[bright60->device[bright60]]% roles
[bright60->device[bright60]->roles]% assign sgeserver
[bright60->device*[bright60*]->roles*[sgeserver*]]% commit


  • Enable the sgeclient roles on the nodes where the jobs are to run. For example, using cmsh:

    [root@bright60 ~]# cmsh
    [bright60]% device
    [bright60->device]% use node001
    [bright60->device[node001]]% roles
    [bright60->device[node001]->roles]% assign sgeclient
    [bright60->device*[node001*]->roles*[sgeclient*]]% commit
    [bright60->device[node001]->roles[sgeclient]]% ..
    [bright60->device[node001]->roles]% ..
    [bright60->device[node001]]% imageupdate -w

After these steps you can control and monitor GE queues and jobs in cmsh and cmgui, and also in the User portal. 


Note, that if CMDaemon is not executing on the node where sge_qmaster is running, then you cannot use the prologue/epilogue scripts distributed with BCM.

Tags: uge sge

Related entries:

You cannot comment on this entry