I am trying to simulate node reboot events in maui's simulation mode.
When node reboot event come, I use "mjobctl -R" to requeue the
jobs on the node, and then use "mnodectl modify state=Down" to
mark the node "Down". And after a few minutes, I use "mnodectl
modify state=Idle" to mark the node "Idle" as the node is available
again.
But this method does not work.
The following error is encountered:
checking node node8
State: Idle (in current state for 00:05:40)
Expected State: Running SyncDeadline: Sat Oct 24 20:26:40
Configured Resources: PROCS: 2 MEM: 1024M SWAP: 2048M DISK: 3000M
Utilized Resources: PROCS: 1
Dedicated Resources: PROCS: 1
Opsys: RedhatAS3 Arch: Xeon
Speed: 1.00 Load: 2.000
Network: [ethernet]
Features: [compute][node8]
Attributes: [Batch]
Classes: [batch 1:2]
Total Time: 00:19:20 Up: 00:18:50 (97.41%) Active: 00:00:20 (1.72%)
Reservations:
Job 'SJ009'(x1) -00:12:30 -> 00:03:12 (00:15:42)
JobList: SJ009
ALERT: jobs active on node but state is Idle
ALERT: node is in state Idle but load is high (2.000)
It seems the node which has been marked Down will get a wrong
state after it is marked Idle.
Can anyone tells me how to solve this problem?
Thanks very much.
--------------
vesor
2006-12-30
_______________________________________________
mauiusers mailing list
[email protected]
http://www.supercluster.org/mailman/listinfo/mauiusers
