Top and other procceses some times hang.with HPUX 11.31

[Wolfgang Belgardt]

Hello,

my customer has a rx7640 system which is configured as 6 Vpars
OS is 11:31
Sometimes the complete system (vpar) does hang ~ 1 minute when starting "top" or other commands
After 1 minute als things runs fine again
in top i can see then a load average = 26,92.
This occured on all partitions on this system
I can't find the root cause of this.
Does someone known this phenomenon?

Thanks in advance

Best Regards

Wolfgang

[Steven E. Protter]

Shalom Wolfgang,

I've seen something similar.


root 1661 1 0 Feb 5 ? 5:07 sendmail: rejecting connections on daemon MSA: load average: 27

Your system has lots of processes waiting.

This is normally a problem with a misbehaving application. Check what you are running and see if someone has some bad code or a known fault with the applications support center.

SEP

[Wolfgang Belgardt]

Hi Steven,


Thanks for your response
i think also, but the problem is to find out which software is the causer of that.
It is sendmail? sendmail has a load of ~ 27

Any Idea?

Best Regards

Wolfgang

[Wolfgang Belgardt]

Hi


an other question is:

is it a generally problem in 11.31 or only a application problem?



Wolfgang

[Torsten.]

In a vPar environment the servers firmware, the vPar software and the entire OS must work closely together.
Especially if you see unspecific problems it is likely a good idea to check if there are newer software, firmware or patches.

[Roland Piette]

Hi Wolfgang,

I have the same problem. Every day I excetute a find command to find out core file at 6 AM. Sometimes this command take 1 minute and on the other time it takes 45 minutes. I installed all latest patches (QPKBASE, Feature 11.i, HWEnable & a Proactive Bundle buid by HP). I remain with this phenomenon!

If a have more news, I will inform you!

Have a nice weekend,
Roland

[Roland Piette · Accepted solution]

Hi Wolfgang,

I have, as promise, another info.

Change the two following tunable parameters in the kernel with the following command :

# kctune vxfs_ifree_timelaq=-1
# kctune vx_ninode=64kb

I had some hanging process and since I change this, my problem is gone. For example : with find, I scan all my local filesystems to find out a specific file. This process can take more than 30 minutes, now it takes 1 minutes !

These parameters are type immediate and can be applied online.

I hope it will help !

Regards,
Roland

[Roland Piette]

Sorry,

Read vxfs_ifree_timelag in place of vxfs_ifree_timelaq !

Roland

[Wolfgang Belgardt]

Hi Roland,


many thanks for your informations about the two parameters.

I will try it.

Regards

Wolfgang

[Bill Hassell]

Are you using SAN disks attached to the computer? Are they older models (ie, Clarion or Symmetrix or similar)? Some of the older arrays have outrageous delays internally (dozens of seconds) to rearrange data or other housekeeping. This is one of the reasons that LVM timeouts are recommended at 60, 90 or even 180 seconds. During that time, all disk I/O is queued and CPU time will also be queued until I/O starts completing. See if your disk array has details about activity at the time you saw delays.