[gridengine users] qhost reports removed compute nodes

Gowtham sgowtham at mtu.edu
Thu Nov 22 16:14:43 UTC 2012


'qhost' on one of of our clusters (running Rocks 5.4.2
with SGE 6.2u5) had been, for a while, reporting nodes
that had long been removed.

   1. 'rocks list host' did not report them
   2. 'ganglia' web interface did not report them


The following attempts did not do the trick:

   1. 'insert-ethers --remove compute-X-Y'
      (this reported that 'compute-X-Y' was removed)
      followed by 'rocks sync config'

   2. Restarting 'gmetad' and 'gmond'


I did steps below to get rid of them from being reported:

   1. qmon
   2. Task --> Host Configuration
   3. 'Delete' the removed nodes from under 'Administration
      Host' tab
   4. 'Delete' the removed nodes from under 'Execution Host'
      tab --> Done
   5. File --> Exit

'qhost' no longer reports the removed nodes.


Best,
g

PS: If this is not recommended approach, please do let me know.

--
Gowtham
Information Technology Services
Michigan Technological University

(906) 487/3593
http://www.it.mtu.edu/




More information about the users mailing list