IBM AIX Powerpath Hdisk removal issue

Question

I have an IBM box running AIX 5.1 ML9 and Powerpath 4.3.0. Recently we migrated from a Symm to a CX700. All went well except this unit now reports 2 dead paths after a reboot (weekly...don't ask!) If I do a "powermt display dev=all" I get (partial display):
---------------- Host --------------- - Stor - -- I/O Path - -- Stats ---
### HW Path I/O Paths Interf. Mode State Q-IOs Errors
==============================================================================
0 fscsi0 hdisk135 SP B3 active dead 0 0
0 fscsi0 hdisk2 SP A2 active alive 0 0
1 fscsi1 hdisk68 SP B2 active alive 0 0

Listing two devices, hdisk135 and hdisk6 (not shown above) as dead. These are not drives currently in the Clariion Storage Groups for this host and do not represent any local drives. We just want to get rid of them.

If I then do:
#powermt check
Warning: CLARiiON device path hdisk135 is currently dead.
Do you want to remove it (y/n/a/q)?

If I answer y(es) it corrects the powermt display, until the next reboot when the 2 dead hdisks reappear. I think I've got an errant entry in my ODM but am hesitant to start cutting on a working system.

Anyone have any experience and thoughts in this area?

TIA Harold Hass

sysmgr1 · Answer

After the 'powermt check', did you try a 'powermt save' ?

Allen Ward · Answer

This sounds something like a situation we have been dealing with, but with different results.

If you could check the following settings...

On the CLARiiON array can you check the initiators for this host and see what the "ArrayCommPath" setting is? Check all the initiators and make sure they are all the same.

In NaviSphere Manager, look at the properties of the host, then switch to the Storage tab and scroll over until you can see the Host LUN ID field. Is there any LUN listed there with a Host LUN ID of 0?

JasonBailey · Answer

After the 'powermt check', did you try a 'powermtsave' ? good question, it would match the symptom

FresnoRed · Answer

Allen,Thanks for the reply but I'm confused. As the end user, my main access to the Clariion is NaviSphere so I don't normally get directly into Clariion. The CE is suppose to be in today so I will ask him to look.In NaviSphere, yes there is a Host LUN ID of 0 but so is there a 0 on all the hosts.Thanks,Harold

FresnoRed · Answer

sysmgr, Jason,Yep, just did the powermt check then save. Will have to wait until next scheduled reboot to see if changes anything. Will post result.Thanks,Harold

FresnoRed · Answer

After doing the powermt save, and after the schedule reboot, I still have the same condition with 2 dead paths:
---------------- Host --------------- - Stor - -- I/O Path - -- Stats ---
### HW Path I/O Paths Interf. Mode State Q-IOs Errors
==============================================================================
0 fscsi0 hdisk135 SP B3 active dead 0 0
0 fscsi0 hdisk2 SP A2 active alive 0 0
1 fscsi1 hdisk68 SP B2 active alive 0 0
1 fscsi1 hdisk92 SP A3 active alive 0 0
0 fscsi0 hdisk9 SP B3 active alive 0 0

and:

### HW Path I/O Paths Interf. Mode State Q-IOs Errors
==============================================================================
0 fscsi0 hdisk10 SP B3 active alive 0 0
0 fscsi0 hdisk5 SP A2 active alive 0 0
0 fscsi0 hdisk6 SP B3 active dead 0 0
1 fscsi1 hdisk7 SP B2 active alive 0 0
1 fscsi1 hdisk8 SP A3 active alive 0 0

When I do a powermt check, I remove the dead hdisks and all looks good again.

Anyone have an idea where I go next?

TIA
Harold

sysmgr1 · Answer

I just removed a number of powerpath devices on one of my AIX machines and they did not return after a reboot. However, all my disks were on a DMX and not Clariion.

The steps I took were ...
- umount the filesystem
- varyoffvg the volume groups
- exportvg the volume groups
- unmapped the disks on the sym so the host wont see them
- cfgmgr
- powermt display showed them as dead paths
- rmdev -Rdl hdisk###
- rmdev -Rdl hdiskpower###
- cfgmgr (and now they are gone)
- rebooted and they are all still gone.

The problem might be that there is a corrupted device table somewhere, you may have to call IBM support. If you also open a support ticket with EMC, they can work together with IBM to help.

JasonBailey · Answer

Anyone have an idea where I go next?

I think there may be problem with the ODM, we have an ODM cleanup procedure in the knowledgebase, see article emc93300. It is an offline procedure needing downtime.

If emc93300 doesnt do it, try emc108127 which has some odmdelete commands.

Finally, log an EMC case if still having trouble.

PowerPath

IBM AIX Powerpath Hdisk removal issue

Was this post helpful?