insf is slow on some systems but not on others...

[Kevin Lister]

Hi Gang,

I've noticed that using "insf -e -H" on some of my HP systems is very slow while on others it's very fast. On a typical system each insf command takes less than 1 second. On some systems it takes several seconds. This occurs when I'm installing device files for hardware paths coming from my va7400's.

All of my OS's are 11.23. My servers consist of N4000's and A500's. The slow behavior happens on both models. So it isn't OS specific or hardware specific.

Anyone else see this behavior?

Thanks in advance for taking the time to answer. ;o)

kev

[Mark McDonald_2]

Are the slow boxes heavily utilised by databases or apps?

[Kevin Lister]

Nope. This problem is not I/O performance related. The boxes are essentially quiescent as I am in the process of setting up storage on them. So, nobody is using them but me when I'm setting them up.

kev

[Avinash20]

Hi Kevin,

Is ioscan is having any devices in NO_HW, we have a known issue

Please check if you are having the patch PHCO_34780 or superseeded one(PHCO_37930) installed on the server

( QX:QXCR1000546202 SR:8606396757 CR:JAGaf56740 )
On IPF platforms, even if disk class devices are already
known to be in NO_HW state (H/W State of devices as seen
with ioscan(1M)), insf(1M) can take a long time to complete.

[Avinash20]

Yo..
"I have assigned points to 0 of 15 responses to my questions."

Please assign point.
If not aware please have a look at
http://forums11.itrc.hp.com/service/forums/helptips.do?#33

Its just a way of saying Thanks !!

[Kevin Lister]

Hi Avinash,

There's no problem with seeing the lun's I created on the va's. I would be trying to install the device files if ioscan failed to see the hardware.

Let me be clear that I'm not having trouble getting disk devices set up. Everything is working. I simply don't see why some of the systems are slow when installing the device files and other systems are not.

It seems quite odd. I through maybe there was a problem with one of the va's. There isn't. I have three va's connected to two different Brocade FC switches. The slow behavior happens through both switches from all three va's on the one system but not on another. Strange.

kev

[Avinash20]

Ok.. so you hav a working and a non-working server

Check with
# timex dd if=/dev/rdsk/ of=/dev/null bs=1024 count=1024

Check if both provide the same output

Also please provide the output of

# timex insf .. ,,the command you are typing

from both working and non-working server.

[Kevin Lister]

Hi Avinash,

All of the servers are working. There are no non-working servers. I've done performance testing on on filesystems after getting them created. They are fine. And before anyone asks... No! It's not a question of using the redundant path to the arrays as opposed to the performance paths. It doesn't matter which path or which array. I can time the execution of the insf command but I don't see any value in that. It takes 5 or so to run. Normal execution time is way less than 1 second. That is what I would expect. I can check the patch levels but I typically perform the exact same install on all servers. I do have a web page I use to compare patch levels on different systems. I wrote a Perl cgi-script that grabs the patch info from each system and then renders the comparison in a simple format that makes any differences stand out very clearly. I'll check that out tomorrow. For now I'm hitting the hay! Talk to you tomorrow! ;o)

kev

[Robert-Jan Goossens]

Kevin,

Could you compare the /etc/loadmods file on your systems, or compare the output of the "kmadmin -s" command and look for the dynamically configurable loadable module krm.

Regards,
Robert-Jan

[Kevin Lister]

Hi RJ,

I'd love to provide you with the info you asked for but I don't have the kmadmin command or the loadmods file on my systems.

Those must be from a different rev of the OS.

kev