Problem with removing a Fiber adapter on VIOserver with MPIO

[AIXtotheworld]

I have a VIOserver with two Fiber adapters, connected to two IBM 2005-16
SAN switch, and a DS8300 with all the LUN’s.
and a bunch of LPAR using more than 150 disks thru the VIO-server.

Maybe I must have “fool-around” with Fiber cables short after the installation, but it have run fine for more than a year. But now the one Fiber adapter is sick.
There was a lot off old “Missing path’s”, but they are removed now.

In the VIOserver I have a defective FC-adapter, I want to remove. (fcs1).
(Is writing in the errlog every second ADAPTER FAILURE)
I have removed all path’es to fscsi1.
I have removed the Fiber cable.
All path’es to the all the hdisk’s is on fscsi0 and working fine!(lspath)

BUT when I try to use hotplug tool to remove the fcs1 adapter I get the messages
“No possible - it is busy.”
I have triyed “rmdev –l fcs1 –R” and “rmdev –l fscsi1”, and get the same error.
When I look on the hdisk’s (lsdev –Ccdisk) can I see that some disks have the same locationcode
as the defective adapter. (and some are sitting on the good one).
Is that the reason to the messages about protocol (fscsi1) is busy ??
Does any buddy know how to fix this situation, without rebooting the VIOserver ?
When I run “fuser –Vxuc /dev/fscsi1”, there are more than 30 process on the entry.

If I power everything down and remove all the hdisk, and run “cfgdev” will the vhost to backing-device mapping go bad?
How can I bee sure that the LUN’s will get the old hdisk number back ?

I hope that some one have some good ideas.

 

[p5wizard]

On VIOS, are you running MPIO with sddpcm or without?

if sddpcm:

step 1:
pcmpath query adapter

(look for your failed adapter name and note the number)

step 2:
pcmpath set adapter X offline
(or)
pcmpath remove adapter X

(X is number from step 1. I'm not at a p5/p6 machine right now so I can't verify man page for pcmpath - AND MAKE SURE you pick the right adapter!)

step 3:
remove/replace adapter with hot plug task on VIOS (oem_setup_env to get to root first)

step 4:
change hostport definition on DS8K so that VIOS LUNs are presented to the new adapter's WWNN.

step 5:
run cfgmgr or cfgdev on VIOS to add redundant paths to the MPIO disks.


HTH,

p5wizard

 

[AIXtotheworld]

Hey p5Wizard
Thank you for your respond !
I have tried your advise with “pcmpath set adapter 1 offline”.
After that all the path’s (“pcmpath query device”) went offline.
But I still have the problem when I try to “remove” the adapter (smitty-diag)
Still same error when making “rmdev –l fcs1 –R” ( or rmdev –l fscsi1 –R).
I don’t understand why, when all the path’s is offline ?
It’s looks like something still is having a connection thru the adapter (og the protocol) ?
Can I remove the OFFLINE paths in some way ?

Regards....

 

[p5wizard]

did you also run this:

[tt]pcmpath remove adapter 1[/tt]


Also, perhaps you need to stop pcmsrv process while doing the hot plug task:

[tt]stopsrc -s pcmsrv[/tt]

(hot plug procedure)

[tt]startsrc -s pcmsrv[/tt]




HTH,

p5wizard

 

[victorv]

hi,
I am not able to work with LPAR and VIOS, but I have some esperience with AIX and FC.

When you give the command "rmdev -l yyyx -R", you ask to the system to remove recursively all the devices that are children of such device: if you remove recurs. fcs1 (the adapter), you will remove the fscsi1 (the protocol) and if there are, the darX (RouterSAN), dacX (the SAN controller),
and .... the disks. If these disks belong to a VG, I belive
you receive a stop to the deleting-chain.

If these VGs are without LVs, remove them before, else, if there are data (and applications) on them, use rmdev (without -R) starting from child (the dacX) until the HBA (the fcsX)


bye