Thank you Gary and all who contributed to this thread. Glad to know this has
been fixed in the next release. I was amazed to see the prompt and quick
response from you guys. Once again thanks much for your time..
This message posted from opensolaris.org
Hi Jeff,
I tried the steps you have suggested. Looks like I am still doing some thing
wrong. Local zone is still unable to get its max assigned cpu when the load
hits 100 %..
Here is the output. Thanks.
bash-3.00# poolstat -r all
id pool type rid rset min
Ramesh, I think that Solaris is not shifting CPUs because the 'used' value is
not large enough to cause a CPU to shift. With the configuration shown below,
a CPU should be shifted automatically when the 'used' value for
rwc-uc-sparc1-z1 exceeds 8.8 (0.8 * 11).
Would you run your test
Jeff, Here is the sample output. I dont see used value is changing at all
even with the load 900
Thanks again.
bash-3.00# poolstat -r all
id pool type rid rset min max size used load
1 rwc-uc-sparc1-z1 pset 1 rwc-uc-sparc1-z14 16 10 0.00
Hi Ramesh,
It looks like you could be falling foul of bug:
6503812 poold not rebalancing very well
http://bugs.opensolaris.org/view_bug.do?bug_id=6503812
http://bugs.opensolaris.org/view_bug.do?bug_id=2146063
This was fixed in snv_56 and back-ported to S10U4.
I believe that you wrote you were
Hi Jeff,
I have MAX_CPUs set to 100.
From your email what I understand is, FSS will work only when multiple zones
using the same pool ?
What I was trying to do is, create a unique pool for each zone and set min and
max pset values and expect for max CPU allocation when load is too high on a
Ramesh Mudradi wrote:
Hi Jeff,
I have MAX_CPUs set to 100.
From your email what I understand is, FSS will work only when multiple
zones using the same pool ?
Yes: setting a value for cpu-shares for a zone will only work:
* if FSS is enabled (which happens by default in OpenSolaris now)
* if
Yes. poold is running.
bash-3.00# svcs -a | grep pool
online 22:32:46 svc:/system/pools:default
online 22:32:56 svc:/system/pools/dynamic:default
bash-3.00# poolstat
pset
id pool size used load
1 rwc-uc-sparc1-z1 11 0.00 101
There doesn't appear to be an objective set for poold. At a minimum you
should read the libpool manpage (man libpool(3LIB)) and understand
the section on setting system.poold.objectives. I would suggest that for
this scenario you should apply the wt-load objective.
Gary
On Thu, May 17, 2007 at
Hi Steve,
After modifying the pset , rebooted the system and then ran the CPU hog script.
So I think pooladm had enough time to read the configuration. Here is the
pooladm output.
# pooladm
system rwc-uc-sparc1
string system.comment
int system.version 1
boolean
10 matches
Mail list logo
