vpars with wlm, cannot move unbound CPU?

[Radoslav Kalaydzhiyski]

As it was in the manual
I used a copy of the the files:
/opt/wlm/examples/wlmconf/vpar_usage_goal.wlmpar
/opt/wlm/examples/wlmconf/vpar_usage_goal.wlm

to start the wlmpard and wlmd

wlminfo slo -l -v
shows:

New messages:
12/01/05 15:46:00 [I] (p19631) wlmd 19631 shutting down
12/01/05 15:46:00 [I] (p29052) wlmd terminated (by request)

New messages:
12/01/05 15:46:04 [I] (p29059) wlmd initial command line:
12/01/05 15:46:04 [I] (p29059) argv[0]=wlmd
12/01/05 15:46:04 [I] (p29059) argv[1]=-A
12/01/05 15:46:04 [I] (p29059) what: @(#)HP-UX WLM A.03.00.01 (2005_08_11_16_50_11) hpux_pa
12/01/05 15:46:04 [I] (p29059) dir: /opt/wlm
12/01/05 15:46:04 [I] (p29059) SLO file path: /var/opt/wlm/.wlmd.cfg.CKP
12/01/05 15:46:04 [I] (p29059) Using absolute CPU units due to primary_host.

New messages:
12/01/05 15:46:04 [W] (p29059) Instant Capacity version not recognized.
12/01/05 15:46:04 [I] (p29059) wlmd 29059 starting
12/01/05 15:46:04 [W] (p29062) Instant Capacity command failures seen while gathering partition status information.

Thu Dec 1 15:46:04 2005

SLO Name Group Pri Metric Goal Req Shares State Concern
slo_myslo - 1 - 10 50.5 100 PASS

Thu Dec 1 15:46:09 2005

SLO Name Group Pri Metric Goal Req Shares State Concern
slo_myslo - 1 - 10 137.98 100 FAIL Priority

Thu Dec 1 15:46:14 2005

SLO Name Group Pri Metric Goal Req Shares State Concern
slo_myslo - 1 - 10 187.24 100 FAIL Priority

Thu Dec 1 15:46:19 2005

SLO Name Group Pri Metric Goal Req Shares State Concern
slo_myslo - 1 - 10 187.29 100 FAIL Priority

Thu Dec 1 15:46:24 2005

SLO Name Group Pri Metric Goal Req Shares State Concern
slo_myslo - 1 - 10 187.66 100 FAIL Priority

#:/homeroot #icod_stat
Software version: B.07.01
System ID: isom2
Serial number: DEH453598S
Product number: A6912A
Unique ID: Z3e116c7516e412ac

Processor Information for Local Virtual Partition
-------------------------------------------------
Assigned processors: 6
Active assigned processors: 6
Inactive assigned processors: 0

Local Hard Partition Information
--------------------------------
Total processors: 8
Active processors: 7

Global iCOD Information
-----------------------
iCOD cells: 0
iCOD processors: 0
iCOD Memory: 0.0 GB


#:/homeroot #vparstatus
[Virtual Partition]
Boot
Virtual Partition Name State Attributes Kernel Path Opts
============================== ===== ============ ======================= =====
TEST1 Up Dyn,Auto,Nsr /stand/vmunix
TEST2 Up Dyn,Auto,Nsr /stand/vmunix

[Virtual Partition Resource Summary]
CPU Num Memory (MB)
CPU Bound/ IO # Ranges/
Virtual Partition Name Min/Max Unbound devs Total MB Total MB
============================== ================ ==== ====================
TEST1 1/ 8 1 0 8 0/ 0 2048
TEST2 6/ 8 6 0 8 0/ 0 12288


Can anybody help me. Why cannot be assigned the unused CPU to this vpar?

[Jonathan Fears]

WLM allows the automatic movement of unbound processors from one vpar to another. From the vparstatus output you've shown, neither vpar has any unbound CPUs. Also from the icod_stat information, the hard partition only has 7 active CPUs. That one unused CPU that you are referring to cannot be used by WLM until it is made active. To see WLM migrate CPUs between vpars, you need to either activate that 1 CPU (then there only be that one which can move from vpar to vpar) or change the vpar setup to have some of those bound CPUs become unbound for the two vpars. Hope this helps.

[Radoslav Kalaydzhiyski]

I added the free processor to the vpar,
but it still cannot be moved on HP-UX 11.11?

[Virtual Partition Resource Summary]
CPU Num Memory (MB)
CPU Bound/ IO # Ranges/
Virtual Partition Name Min/Max Unbound devs Total MB Total MB
============================== ================ ==== ====================
TEST1 1/ 8 1 0 8 0/ 0 2048
TEST 6/ 8 6 1 8 0/ 0 12288




Mon Dec 5 10:37:09 2005

SLO Name Group Pri Metric Goal Req Shares State Concern
slo_myslo - 3 - 16 339.38 700 PASS

Mon Dec 5 10:37:15 2005

SLO Name Group Pri Metric Goal Req Shares State Concern
slo_myslo - 3 - 16 339.26 700 PASS

Mon Dec 5 10:37:19 2005

SLO Name Group Pri Metric Goal Req Shares State Concern
slo_myslo - 3 - 16 339.06 700 PASS


SLO Name Group Pri Metric Goal Req Shares State Concern
slo_myslo - 1 - 16 184 100 FAIL Priority

Mon Dec 5 10:37:36 2005

SLO Name Group Pri Metric Goal Req Shares State Concern
slo_myslo - 1 - 16 184 100 FAIL Priority

Mon Dec 5 10:37:41 2005

SLO Name Group Pri Metric Goal Req Shares State Concern
slo_myslo - 1 - 16 184 100 FAIL Priority

Mon Dec 5 10:37:46 2005

SLO Name Group Pri Metric Goal Req Shares State Concern
slo_myslo - 1 - 16 184 100 FAIL Priority


The most interesting is that it works on HP-UX 11.23 we have another npar with vpars and HP-UX 11.23.
But this is on HP-UX 11.11, and there the WLM seems not work as expected.

[Jonathan Fears]

Hello again...When you activated that last CPU, did you restart WLM? If not, it may be that WLM is still only working with the 7 CPUs that you had active at the time it was started. I'm sure you thought of this also, but make sure the SLOs that are failing are on the partition that does not have the unbound CPU. Are there any other differences between your 11.23 and 11.11 systems?
If you want to send an email to wlmfeedback@rsn.hp.com, I would be happy to continue helping you figure out what is going on with your setup.

[Radoslav Kalaydzhiyski]

The problem was in icod version, so add this line "icod:3:7.01" in "/opt/wlm/config/versions"