Very Computer

by **Yakov Yukhnovetsk** » Fri, 27 Feb 2004 02:04:21

Hi !
Does anybody care, that last Recommended Patch Cluster for Solaris 8 is
buggy ?
I'll try to be more precise:
We have several different Sun computers - U60, R280, V250 and so on.
For all these architectures we use to install the same patch level.
We do it from time to time. The last time we installed Recommended Patch
Cluster from 16.01.04 (patch level Generic_108528-27).
Two strange things :
First, tcsh is wrecked. OK, it is not, actually, a problem. I just copy
tcsh from another machine over new tcsh.
Second - our application while running in absolutely "sleepy" mode,
leads to huge load average.
Look at the top's output:

load averages: 15.42, 15.63, 14.62
18:47:43
66 processes: 65 sleeping, 1 on cpu
CPU states: 96.4 % idle, 1.1 % user, 0.8% kernel, % iowait,
% swap
Memory: 1024M real, 174M free, 647M swap in use, 4239M swap free

PID USERNAME THR PRI NICE SIZE RES STATE TIME CPU COMMAND
890 corems 13 59 0 64M 48M sleep 0:08 0.40% SubnetworkMgr
889 corems 14 58 0 51M 35M sleep 0:07 0.21% TrapManager
1031 corems 1 48 0 1824K 1176K cpu 0:00 0.19% top
902 corems 8 55 0 48M 32M sleep 0:04 0.12% RTDiscoverer
901 corems 9 58 0 31M 26M sleep 0:04 0.11%
EventDispatcher
897 corems 10 58 0 52M 38M sleep 0:05 0.11% AlarmManager
906 corems 12 58 0 70M 55M sleep 0:09 0.09% DomainManager
903 corems 13 58 0 71M 57M sleep 0:15 0.06% DomainManager
322 corems 7 58 0 26M 22M sleep 0:05 0.02% itnode_daemon
886 corems 8 58 0 50M 37M sleep 0:06 0.02% LogService
900 corems 7 58 0 48M 32M sleep 0:03 0.02% PerfMgrm
374 oracle 1 58 0 385M 339M sleep 0:00 0.02% oracle
894 corems 9 58 0 50M 36M sleep 0:03 0.01% ProfManager
576 oracle 1 48 0 385M 342M sleep 0:00 0.01% oracle
327 corems 11 58 0 38M 31M sleep 0:18 0.00% itnotify

NOTHING is on CPU, IDLE is almost 100%, but laod average is huge. Can
somebody explain ?

The same application, I mean the SAME binaries, while running on
another computer with patch level Generic_108528-22, for example, produces
load average less than 1.
(In case of potention question - yes, we recompiled all the our executables
on computer with new patch level - the same result.)
OK, today I've installed the very new recommended patch cluster
(Generic_108528-29, from 20.02.2004) - the same problems.

BTW, top is OK, sdtperfmeter shows the same load average, as top.

Somebody from Sun knows about these "oddities" ?

Thanks

Corrigent Systems.

by **Bart Somer** » Fri, 27 Feb 2004 04:22:36

Actually, it's not.

You should be able to see what's going on using iostat and prstat. I
guess it's trying to release some lib's which are in use before updating
them. And also you could get some updates for your ALOM and probably
QUS-scsi-drivers. This is the reason Sun recommends to run the
8_Recommended in single-user mode.
Personally once i had problems with corrupted passwd files after running
a patchcluster in full operational mode.

HTH,

Bart

> Hi !
> Does anybody care, that last Recommended Patch Cluster for Solaris 8 is
> buggy ?
> I'll try to be more precise:
> We have several different Sun computers - U60, R280, V250 and so on.
> For all these architectures we use to install the same patch level.
> We do it from time to time. The last time we installed Recommended Patch
> Cluster from 16.01.04 (patch level Generic_108528-27).
> Two strange things :
> First, tcsh is wrecked. OK, it is not, actually, a problem. I just copy
> tcsh from another machine over new tcsh.
> Second - our application while running in absolutely "sleepy" mode,
> leads to huge load average.
> Look at the top's output:

> load averages: 15.42, 15.63, 14.62
> 18:47:43
> 66 processes: 65 sleeping, 1 on cpu
> CPU states: 96.4 % idle, 1.1 % user, 0.8% kernel, % iowait,
> % swap
> Memory: 1024M real, 174M free, 647M swap in use, 4239M swap free

> PID USERNAME THR PRI NICE SIZE RES STATE TIME CPU COMMAND
> 890 corems 13 59 0 64M 48M sleep 0:08 0.40% SubnetworkMgr
> 889 corems 14 58 0 51M 35M sleep 0:07 0.21% TrapManager
> 1031 corems 1 48 0 1824K 1176K cpu 0:00 0.19% top
> 902 corems 8 55 0 48M 32M sleep 0:04 0.12% RTDiscoverer
> 901 corems 9 58 0 31M 26M sleep 0:04 0.11%
> EventDispatcher
> 897 corems 10 58 0 52M 38M sleep 0:05 0.11% AlarmManager
> 906 corems 12 58 0 70M 55M sleep 0:09 0.09% DomainManager
> 903 corems 13 58 0 71M 57M sleep 0:15 0.06% DomainManager
> 322 corems 7 58 0 26M 22M sleep 0:05 0.02% itnode_daemon
> 886 corems 8 58 0 50M 37M sleep 0:06 0.02% LogService
> 900 corems 7 58 0 48M 32M sleep 0:03 0.02% PerfMgrm
> 374 oracle 1 58 0 385M 339M sleep 0:00 0.02% oracle
> 894 corems 9 58 0 50M 36M sleep 0:03 0.01% ProfManager
> 576 oracle 1 48 0 385M 342M sleep 0:00 0.01% oracle
> 327 corems 11 58 0 38M 31M sleep 0:18 0.00% itnotify

> NOTHING is on CPU, IDLE is almost 100%, but laod average is huge. Can
> somebody explain ?

> The same application, I mean the SAME binaries, while running on
> another computer with patch level Generic_108528-22, for example, produces
> load average less than 1.
> (In case of potention question - yes, we recompiled all the our executables
> on computer with new patch level - the same result.)
> OK, today I've installed the very new recommended patch cluster
> (Generic_108528-29, from 20.02.2004) - the same problems.

> BTW, top is OK, sdtperfmeter shows the same load average, as top.

> Somebody from Sun knows about these "oddities" ?

> Thanks

> Corrigent Systems.

by **Raf LaPiet** » Fri, 27 Feb 2004 04:31:51

> First, tcsh is wrecked. OK, it is not, actually, a problem. I just copy
>tcsh from another machine over new tcsh.

If by "tcsh is wrecked" you mean tcsh file completion shows "0c" instead of
a file name, this is a known bug (4961642) and you can fix it by backing out
patch 110943-02 or by calling sun and installing the T-patch which fixed the
problem

HTH

-Raf

by **Yakov Yukhnovetsk** » Fri, 27 Feb 2004 18:18:02

Ha-ha-ha !
Probably I I do not understand. If in my work I knew about a stupid bug but
did not fix it, probubly I 'd got fired.
Why not to fix it in the nearest patch cluster ?

Do you know? probably, something about load average ?
Thanks.

> > First, tcsh is wrecked. OK, it is not, actually, a problem. I just
copy
> >tcsh from another machine over new tcsh.
> If by "tcsh is wrecked" you mean tcsh file completion shows "0c" instead
of
> a file name, this is a known bug (4961642) and you can fix it by backing
out
> patch 110943-02 or by calling sun and installing the T-patch which fixed
the
> problem

> HTH

> -Raf

Very Computer

Recommended Patch Cluster for Solaris8 is BUGGY !!!!

Recommended Patch Cluster for Solaris8 is BUGGY !!!!

Recommended Patch Cluster for Solaris8 is BUGGY !!!!

Recommended Patch Cluster for Solaris8 is BUGGY !!!!

Recommended Patch Cluster for Solaris8 is BUGGY !!!!