N-Class with A5158A and EMC Clariion performance problem

[john_301]

Dear All,
I'm facing this intresting problem.
old configuration:
N-class machine connected to HP jamaica disk array.

New configuration:

Add A5158A adapter and EMC clariion FC4700 disk array.
After upgrading to fiber channel EMC disk array the machine become very slow and Glance showed the CPU time 100%!!

nothing changed in the application just creating new volumes in EMC and mount it instead of the old volumes! EMC clariion had no errors.

with the old configuration we don't had any performance problems!.

The Swap % and MEM % from Glance is almost the same before and after upgrade.

The driver ver. of the A5158A is B.11.00.10 and the system is running HPUX 11.0 and upgraded to the lates patches.

Did anybody face this problem before?
any help pls.

[Eugeny Brychkov]

Can you see from Glance which process eats 100% of CPU time? I guess that is something wrong with configuration, and it makes FC adapter I/O queue full - so you have slow responce and high CPU usage. Did you set EMC up to work in hpux mode (sorry, not familiar with EMC, but HP disk arrays should be said to work in certain OS mode to support that OS)? Do you use FC switch? If yes, do you have everything in fabric?
Check if FC card works in fabric:
- tdutil /dev/tdX
- ioscan -fn
attach both outputs to your next reply
Eugeny

[Martha Mueller]

We have had the same configuration as you....going from jamaica's to an fc4700 with an N4000. The only time we had the high cpu usage was after we had a disk fail in the fc4700. There was a bug in the version of software on the fc4700 that couldn't handle having two hot spares available. It kept rebuilding onto the hot spare, over and over again. I think we let it rebuild about 4 times. We ended up having to rebuild that raid group, so all the data in that raid group was destroyed. We had time to save it to tape first, but that server was down for 9 hours while we rebuilt the raid group and then restored from tape. The current software doesn't have that bug anymore.

[john_301]

Thanks for fast reply,


The EMC box is connected directly to the N-Class server. and the Application is the one having the maximum CPU %.
The EMC Status is OK there isn't any fault. From Glance the HDD usage is 20% (EMC box).

I'm not expecting the application because it is working fine with the old disk array

Martha,
Regarding the EMC S/W what ver. you are using?

[Sridhar Bhaskarla]

Hi,

Can you post a snippet of the following commmands?

sar -d 2 5
sar 2 5

Thanks,
Sridhar

[john_301]

Really I can't issue any command because The machine is in production and I'm using now the old configuration, it needs arrangement to reactivate the new configuration. Really I used sar before and the idl % was 0.

[Martha Mueller]

Base 08_43_57

[Martha Mueller]

08_43_57 is a version without the bug I mentioned.