Opened 12 years ago

Last modified 9 years ago

#466 new defect

IZ2390: queues with 0 slots suspends subordinate

Reported by: sgaure Owned by:
Priority: lowest Milestone:
Component: sge Version: 6.1u2
Severity: minor Keywords: qmaster
Cc:

Description (last modified by admin)

[Imported from gridengine issuezilla http://gridengine.sunsource.net/issues/show_bug.cgi?id=2390]

   Issue #: 2390   Platform: All   Reporter: sgaure (sgaure)
   Component: gridengine   OS: All
   Subcomponent: qmaster   Version: 6.1u2   CC: None defined
   Status: NEW   Priority: P5
   Resolution:   Issue type: DEFECT
     Target milestone: ---
   Assigned to: ernst (ernst)
   QA Contact: ernst
   URL:
   * Summary: queues with 0 slots suspends subordinate
   Status whiteboard:
   Attachments:
   Date/filename:                           Description:                                                                                  Submitted by:
   Fri Oct 5 10:31:00 -0700 2007: bad2.rqs  disabled rqs's - output of qconf -srqs immediately after restart of sge_qmaster (text/plain)  sgaure
   Fri Oct 5 10:32:00 -0700 2007: good2.rqs disabled rqs's - output of qconf -srqs immediately before restart of sge_qmaster (text/plain) sgaure
     Issue 2390 blocks:
   Votes for issue 2390:  1

   Opened: Fri Oct 5 02:56:00 -0700 2007 
------------------------


Just a minor issue.

If I create a queue A with slots=0 and subordinate_list B=1, all instances of B
are immediately suspended.  According to the documentation, suspension should
happen if I had 'subordinate_list B', because A is technically "full", but with
'B=1' suspension should not happen.

This popped up because we happened to set slots=0 on queue A to prevent jobs
from starting there temporarily; it did not have the desired effect on queue B
 :-)

   ------- Additional comments from sgaure Fri Oct 5 10:31:26 -0700 2007 -------
Created an attachment (id=128)
disabled rqs's - output of qconf -srqs immediately after restart of sge_qmaster

   ------- Additional comments from sgaure Fri Oct 5 10:32:24 -0700 2007 -------
Created an attachment (id=129)
disabled rqs's - output of qconf -srqs immediately before restart of sge_qmaster

   ------- Additional comments from sgaure Fri Oct 5 10:34:34 -0700 2007 -------
whoops, attached to wrong issue, should be 2387

   ------- Additional comments from jlopez Mon Jan 12 04:46:42 -0700 2009 -------
This bug is a serious problem if you try to implement the "Exclusive Host Access
With Grid Engine" described in:

http://blogs.sun.com/templedf/entry/exclusive_host_access_with_grid

and temporary set slots=0 in one node.

Even if you use queue=1 when the number of slots of in the node is 0 (slots=0)
the queue is treated as if it was full and it suspends its subordinated queues.
This causes the unwanted effect that all queues of the cycle used to implement
the exclusivity feature get suspended. After that they do not get back to the
unsuspended state unless you remove the node from the queue and add it again.

Attachments (2)

128 (4.2 KB) - added by dlove 9 years ago.
129 (4.6 KB) - added by dlove 9 years ago.

Download all attachments as: .zip

Change History (3)

Changed 9 years ago by dlove

Changed 9 years ago by dlove

comment:1 Changed 9 years ago by admin

  • Description modified (diff)
  • Severity set to minor
  • Summary changed from IZ2390: ith 0 slots suspends subordinate to IZ2390: queues with 0 slots suspends subordinate
Note: See TracTickets for help on using tickets.