This post is more than 5 years old
90 Posts
0
3643
Need to replace HBA on AIX box
It's been so long since we had a problem with an HBA, I need guidance on replacing one on one of my older boxes. I have a 6M1 running AIX 5.3.0-05 and Powerpath 3.0.4. One of my IBM 2765 cards stopped working and pp worked so I still have LUNs from my CX700 (2.19.700.5.030.) So I am placing a support call to IBM to replace the card but I need to prep to change the card and reconfigure powerpath. I got a pretty good handle on the zoning and probably can muddle thru the Navisphere connectivity, but am unsure on the AIX stuff. Any help would be appreciated. We are working up a maintenace window for this weekend so I have some time. This is our payroll system, so I really, really, don't want 9000 county employees mad at me.
Thanks in advance,
Harold
County of Fresno
Thanks in advance,
Harold
County of Fresno
emc_troy
142 Posts
0
March 26th, 2009 16:00
# powermt remove hba=X
Go directly to # diag for eplacement. It may complain thet device have to be unconfigured, and on the same level where replace PCI is, there is "unconfigure", or you may use
# rmdev -l fcsX -R
to delete fcs definition & all the child devices - hdisX & FC protocols
Again, after replacing the HBA: do zoning/masking
# emc_cfgmgr or cfgmgr
Get the WWN of the new HBA:
# lscfg -vl fcsX
Do zoning/masking
# emc_cfgmgr or cfgmgr
# powermt config
# powermt display (verify you see the new adapter/disk as part of hdiskpowerX
# powermt save
emc_troy
142 Posts
0
March 26th, 2009 09:00
This is an on-line operation if the IBM HW supports it & you have redundant connections for this server. I don't have AIX to confirm and it's been a while I've done it but these are the steps:
1. Unconfirmed the disk from PP: # powermt delete hba=X , where X is the HBA instabce
2. Once you have the new HBA: # diag -> scsi/on-line replace....follow the instructions
3. Zoning/Masking the old disk to the new HBA
4. # powermt config, # powermt save
5. You are good to go
FresnoRed
90 Posts
0
March 26th, 2009 12:00
Thanks for the response. Question; the diag you refer to, you talking about AIX diag? If so I don't find any reference to "scsi/on line replace." Again thanks for getting back.
Harold
emc_troy
142 Posts
0
March 26th, 2009 13:00
# diag -> "AIX tasks and service aids" (I think it was the second menue) ->Hot plug task -> PCI hot plug manager -> Replace/remove a PCI hot plug adapter
See where this command will position you: # diag -T"identifyRemove"
FresnoRed
90 Posts
0
March 26th, 2009 15:00
Yes, thank you, I don't normally think of these old 6M1's as hot swapable so it didn't look right. Since we are having IBM do the replacement, and IT'S PAYROLL; mgmt. is being way cautious and we will be doing a system shutdown on Sun. morning to swap out the card. So, I need to do the stuff to get ready. In the past we would rmdev the fc card, the hdisks, the hdiskpowers associated with the path. Then we would have a head scratching session trying to figure out what to do next. I'm trying to make that more elegant. Will powermt remove kill out the devices in that path for me? And can I do that while the system is up?
I know this is asking a lot, but, in case I haven't mentioned it, IT'S PAYROLL.
Thanks again,
Harold
FresnoRed
90 Posts
0
March 27th, 2009 07:00
Thanks for your patience. Will be replacing the HBA Sunday and I will post back results.
Harold
FresnoRed
90 Posts
0
March 27th, 2009 07:00
emc_troy
142 Posts
0
March 30th, 2009 09:00
"We bounced the box prior to replacing the card but that didn't help (didn't expect it to but...) " - sometimes it helps. I've seen cases when the HBA just hangs - power it off/on sometimes reset it. It can be done by reboot or with the diag: simulating replacing the Adapter.
But if it happens once ...it will happen again so replacing the HBA is better option.
Looks like I still have some memory from my AIX Admin years
xe2sdc
2.8K Posts
0
March 30th, 2009 09:00
FresnoRed
90 Posts
0
March 30th, 2009 09:00
We bounced the box prior to replacing the card but that didn't help (didn't expect it to but...) Noticed the hba# for fcs0 changed from 0 to 1. This was an important thing to notice since the next step was powermt remove hba=1. This took out the hba and the hdisk and hdiskpower devices were also gone. We expected to have to do some explicit rmdev's but this made life easier. Then did the zoning on the switch and saw the new unregistered wwn's in Navisphere. Removed the old and registered the new paths to the host. Still didn't see the LUNs at the host. Did powermt configs, emc_cfgmgrs, inq.aix64's but still no LUNS. Went back to Navisphere and realized I had neglected to Reconnect. Did that and after one more reboot, all was good.
Hope this posting helps some other lost soul when it comes to doing something that only happens once in 7 years.
Thanks Troy for your help.