IO Timeout and Powerpath

[Peter Marais]

How do I change the IO timeout param with emc powerpath installed. Tried pvchange -t but it comes back with:
pvchange: Physical voulume :/dev/dsk/c4t0d0" could not be changed:
No such device or address.

pvdisplay /dev/dsk/c4t0d0 has no problem in listing the parameters.

[Cheryl Griffin]

This will happen if the pvchange process times out.

Check your Powerpath version. You may have to update it.

[Peter Marais]

The Powerpath version is HP.4.3.0 I beleave it to be pretty resent. Also the pvchange command fails immediately. No delays

[Cheryl Griffin]

Have you made any changes on the EMC and if so have you tried a reboot and the pvchange again? If changes are made, the disk driver may no longer be communicating so the disk looks like it's not there.

If this does not help, someone in the Response Center can help you if you open a call.

They need to analyze an lssf, ioscan, a pvdisplay, diskinfo -v, look at the drivers in the kernel (specifically scsi drivers) and if there are any associated messages in syslog. There are some internal tools they can provide you that can help analyze the situation as well.

[Peter Marais]

What I would like to know is weather anyone tried to change the io timeout paramater while powerpath is installed. I have just been recommended to change the value from default to 90 seconds (by emc grab). The system is behaving normally and I do not want to spend time on this.

[Florian Heigl (new acc)]

can You supply the full command and output, and also a lssf /dev/dsk/c4t0d0 and lssf /dev/rdsk/c4t0d0 ?

To answer Your question: I never tried it in a power path config, for we only use pvlink on our hp-ux boxes. You should clarify with EMC^2 if You really need to set it - in my experience it's crucial in Clariion/CX + pvlink environments and close to irrelevant with PowerPath and i.e. DMX.

OTOH the default of 30 seconds is a bit short for hot-wiring fibre links ;)

[john kingsley]

Yes, I was able to change the timeout value on my physical volumes, and I am running powerpath ver. 4.2. EMC reccomended a value of 180 sec. I'm not sure if it maters, but check to see if you are running pvchange on the primary path.

[Peter Marais]

Thanks everyone for the interest so far. I am attaching a file of what I did.

[john kingsley]

Do you know where the dead path came from?
It can be removed by running:
powermt check dev=path

[OFC_EDM]

Regarding the "no such device or address".

We used to have similar issues when adding LUNS.

But now before we do anything with the EMC disks we do the following:

1. powermt check
--> if it prompts to remove dead paths we remove all.

2. If adding LUNS we create them and assign them to the host at this point via Navisphere.

3. ioscan -fnC disk
4. insf -C disk
5. powermt config
6. powermt save
7. Then we review everything with powermt display dev=all and make sure it matches up with Navisphere.

This is only loosley related to your issue in that we do this to ensure dead paths are removed and that the devices you see with pvdisplay really exist.

Cheers.

[Peter Marais]

Tried removing the unknown path but could not. See attachment

[OFC_EDM]

What's the output from

ioscan -fnC disk

?

[OFC_EDM]

Did you have an HBA replaced recently?

[sathish kannan]

Hi Peter,
Looks to me you replaced a HBA or changed the instance before removing those devices from Powerpath. Have a look into the "ioscan" output and check for H/W path with no hardware or unknown status and the run "powermt check"

Regards
Sathish

[Peter Marais]

We had a path change some time ago where EMC replaced one of there switches for a newer one. This caused the path change and regenned the device files. All the unknown paths and old device files have been removed. Also should the powerpath check command not also remove the unknown device from EMC? How do i repair this the system has been rebooted after cleaning up the device files so a reboot will not cure this. Also the HBA's are functioning fine so no problem there. Rechecked ioscan -fnC and all devices displayed are claimed.

[Bernd Reize]

To remove a dead hba form powerpaths config you have to issue "powermt remove hba=6" manually. (Please verify the controller instance number, before doing this!)

[Peter Marais]

Tried to remove it, no sucksess.

[Milton Lin_3]

If you still see UNKNOWN paths,

couple of things to try....
1.#powermt check
or
#powermt check force
2. Check from the SAN/Navisphere the connectivity status and verifiy the number of paths/connections.
3. #isocan -fnC disk then #powermt config

------------
Timeouts as I have seen vary depending on Clarrion and Symmetrix. I've seen recommendation of 90 seconds for Powerpath and 180 seconds for PVLINKS
------------
good luck

[Peter Marais]

Thanks Lin but I have tried that already.

[sathish kannan · Accepted solution]

Hi Peter,
You need to rebuild your powerpath database.
Take note of your current configuration and rebuild powerpath configuration.

Regards
Sathish

[Peter Marais]

Thanks Sathish, how do I rebuild the database?

[Peter Marais]

Rebuilt and rebooted. This cleared the no device and I could set iot