hwmgr hierarchy errors

[Blake Roberts]

Folks,
I have a situation where I had to replace a failed HBA (KGPSA). When I replaced it, a hwmgr -view hierarchy still only shows my SAN attached disks as being connected to one HBA. They are also flagged as critical. A hwmgr -status component shows them as critical/unavailable, although my filesystems are still mounted. This essentially gives me some reservations about whether or not my I/O is distributed down both paths.

My OS is 5.1B PK3 with FW 6.6. This is a GS1280 system connected to an EVA 5000. Attached is a list of the hwmgr output.

My question is whether or not there is a way that I can get the stale paths deleted from the disks so that both paths will show the devices. Right now, a hwmgr -refresh scsi and a hwmgr -refresh component will not do it for these disks.

Thanks in advance,
--Blake Roberts

[Uwe Zessin · Accepted solution]

Blake,
it sounds like the only other way is to recreate the device, please see his thread:
http://forums1.itrc.hp.com/service/forums/questionanswer.do?threadId=430956

[Johan Brusche]

Blake,

Are you using "selective presentation" on the EVA5000 ? If yes, then the new KGPSA presents a NEW connection to the EVA, so you need to remove the old connection ID of the broken KGPSA and assign the new connection the storage units, else emx3 innode ERPEMSA will never see any devices.

Johan.

[Ralf Puchner]

As I know there is a special "replacing" manual preventing the described problem.

Please open a call within the HP support center and ask for assistent (it would be better to ask before replacing components)

[Blake Roberts]

Uwe confirmed what I was hoping wouldn't be the case. I had a call open with support, and they informed me of the same thing. I have just been living in denial, thinking that there was a better way of doing this.

Also, on the EVA, I am using selective presentation, but I did remove the old connections from the EVA so that I would not have that problem from that level anymore.

Thanks for all of the feedback!
--Blake

[Uwe Zessin]

I have been told today that a 'hwmgr -refresh scsi' is supposed to remove stale paths as long as the first path is not stale.

[Blake Roberts]

Today, I finally got a chance to work on cleaning up the hardware database. I had run hwmgr -refresh scsi and hwmgr -refresh component previously, with no good results.

I rebooted the system so that I could do the work in single-user mode, and when I checked the hierarchy, all of the critical device warnings were gone, and the devices showed up on both paths. I noticed that the stale paths still show up, but the devices are now visible on both paths. A hwmgr -get attr -id xxx shows 2 valid paths, 2 standby paths, and 2 failed paths, instead of 1 valid, 1 standby, and 2 failed.

In a nutshell, the stale paths are there, but the devices now have the visible paths that they are supposed to have.

Thanks for the help,
--Blake Roberts