Replacing fibre channel card without server reboot. 1

[ggitlin]

Hello, AIX gurus!
We have AIX 5.2 p690 LPAR with a failed HBA. We need to replace it and get the fcs1 into the DEFINED state, before the hardware can be replaced. However, there are a lot of child devices on the adapter. Is it possible to down the HBA without removing the disks/vpaths?

 

[p5wizard]

Depends on what version of SDD you are running...

if you have the option
datapath remove adapter
then yes it can be done online. If not, then you need a downtime...

step 0 - identify your faulty adapter name
lsdev -Ccadapter | grep fcs

step 1 - identify your faulty adapter number in SSD config
datapath query adapter

step 2 - remove adapter from SDD config
datapath remove adapter N
(use number from step 1 instead of N)

step 3 - remove adapter from ODM
rmdev -Rdl /dev/fcsM
(use number from step 0 instead of M)

step 4 - use diag/PCI hot plug tasks to replace the adapter - probably your CE will do this.

step 5 - identify your new WWNN and adapt your storage server to assign the LUNS to the new WWNN
lscfg -vl fcsM

step 6 - run cfgmgr to rediscover the deleted paths to your LUNs
cfgmgr -vl fscsiM

step 7 - add paths to SDD
addpaths

step 8 verify all LUNS have same number of active paths
datapath query device
datapath query adapter



HTH,

p5wizard

 

[ggitlin]

Thanks, Wizard, but,
When the hdisks are re-discovered, will they have the same numbers (like hdisk135)? My concern is if they will not, that might prevent access to the disks, as vpaths are associated with specific hdisks.

 

[p5wizard]

No, they might not have the same numbers, but you needn't worry.

datapath remove adapter also removes the concerned hdisks from each vpath - check it with datapath query device after datapath remove adapter.

then afterwards, with newly rediscovered hdisk paths, the addpaths command matches the newly discovered hdisks with their correct siblings. I've tried it, it works.



HTH,

p5wizard