[Mauiusers] Re: Standing reservation seems to apply to more hosts than defined in HOSTLIST

Matthias Lein m.lein at massey.ac.nz
Wed Aug 22 15:26:50 MDT 2007


Hi Group!

I have a peculiar problem here:

I have several nodes called work1, work2, ..., work10, work11, ...
And on node "work1" there is a standing reservation for working hours:

SRCFG[work1]           USERLIST=lein
SRCFG[work1]           OWNER=USER:lein
SRCFG[work1]           HOSTLIST=work1
SRCFG[work1]           PERIOD=DAY DAYS=MON,TUE,WED,THU,FRI
SRCFG[work1]           STARTTIME=8:00:00 ENDTIME=20:00:00
SRCFG[work1]           ACCESS=shared

However, when I check the reservations with showres, it seems that
"work1" is interpreted as part of a hostname - so in effect
reservation work1.0.0 places a stadning reservation for work1, work10,
work11, ...

reservations on Wed Aug 22 10:26:01

           NodeName       Type      ReservationID   JobState Task
 Start    Duration            StartTime

              work1       User          work1.0.0        N/A    1
00:00:00     9:33:59  Wed Aug 22 10:26:01
             work10       User          work1.0.0        N/A    1
00:00:00     9:33:59  Wed Aug 22 10:26:01
             work11       User          work1.0.0        N/A    1
00:00:00     9:33:59  Wed Aug 22 10:26:01
             work12       User          work1.0.0        N/A    1
00:00:00     9:33:59  Wed Aug 22 10:26:01
             work13       User          work1.0.0        N/A    1
00:00:00     9:33:59  Wed Aug 22 10:26:01
[and so on]

Besause of other reservations of the same kind on some of the machines
(work1[0-9]), checknode tells me that the node is overcommitted:

checking node work11

State:      Busy  (in current state for 00:00:00)
Expected State:     Idle   SyncDeadline: Sun Oct 25 01:26:40
Configured Resources: PROCS: 1  MEM: 2012M  SWAP: 2446M  DISK: 89G
Utilized   Resources: PROCS: 1  DISK: 29G
Dedicated  Resources: [NONE]
Opsys:        Ubuntu  Arch:      x86_64
Speed:      1.20  Load:       1.230 (MaxLoad: 0.90) (ProcSpeed: 2400)
Network:    [DEFAULT]
Features:   [NONE]
Attributes: [Batch]
Classes:    [batch 1:1]

Total Time: 00:51:00  Up: 00:51:00 (100.00%)  Active: 00:51:00 (100.00%)

Reservations:
 User 'work1.0.0'(x1)  -00:02:05 -> 9:31:54 (9:33:59)
   Blocked Resources at -00:02:05   Procs: 1/1 (100.00%)
 User 'work11.0.0'(x1)  -00:02:05 -> 9:31:54 (9:33:59)
   Blocked Resources at -00:02:05   Procs: 1/1 (100.00%)
 ALERT:  no jobs active on node for  -INFINITY but state is Busy
ALERT:  node is overcommitted at time -00:02:05 (P: -1)

Then, diagnose -r tells me that there is only "work1" in the hostlist,
but says that there are 5 nodes and 6 processors reserved...

Diagnosing Reservations
ResID                      Type Par   StartTime     EndTime
Duration Node Task Proc
-----                      ---- ---   ---------     -------
-------- ---- ---- ----
work1.0.0                  User des   -00:03:38     9:30:21
9:33:59    5    5    6
   Flags: STANDINGRES
   ACL: RES==work1.0= USER==lein+
   CL:  RES==work1.0
   Owner:  USER:lein
   Task Resources: PROCS: [ALL]
   Attributes (HostList='work1')
   Active PH: 0.00/0.36 (0.00%)
   SRAttributes (TaskCount: 0  StartTime: 8:00:00  EndTime: 20:00:00
Days: Mon,Tue,Wed,Thu,Fri)

Am I missing something here? How can I keep maui from expanding
"work1" into "work1*"?

Cheers,
Matthias


More information about the mauiusers mailing list