[Stef]

Hi,

We have problems with rdac drivers on AIX. It's not the first time that we have problems and untill know, IBM couldn't gave a solution.

What's going on? An AIX server connected to a DS4xxx box. Rdac is running fine. So 2 dac's, 1 dar and failover is working.

We want to put 1 controller of the DS box in an other SAN port. To do so, we remove the dac from the AIX so we only see the controller that's will not change. After putting the controller in the other SAN port and running cfgmgr again, troubles start. It depends from server to server and what steps you do, but basically, the dar remembers the old dac and refuses to connect the new dac to the dar.

So, you get something like dar0 is dac0 and dac1. I remove dac0 switch the DS and dar0 uses still dac0 and dac1, but I also have a dac2. The dac0 is gone, but it still appears when I do "lsattr -El dar0".

I can recover from this by removing dar0 and run cfgmgr. This is not always possible if you boot from the san.

It sounds confusing, but feel free to ask more information

Stef

[moralebr]

Stef,

When you say "we remove the dac from AIX", did you use the command "rmdev" with "-d" parm.?

Recheck if you don't forgot some zoning in SAN switch. If you forgot it, your machine is seeing duplicated zones/controllers.

Regards,
Luciano Martins.

[ScottMcG]

Hi Stef,

RDAC drives me crazy from time to time. Could you post the output of "lsattr -E -l <item>" for each of your dacs and dar? Also, could you comment on how your SAN is zoned? From what you said, I made the assumption that you are connecting via a switch or director rather than direct attaching from the HBAs in your pSeries box to the host ports on your DS4000.

Regards,
Scott

[Stef]

Hi,

(damn, I typed a hugh reply, but I got a server timeout)

For the zoning, 1 zone with each time 1 initiator and 1 target. 2 zones per server so each server has 1 path to controller A and 1 path to controller B.

And no, it's not the first time I have problems with rdac. I can understand a mistake is problem, but it _never_ worked like it should be. As soon as you put a DS an other port, problems are starting.

Basically, these are the steps we did
- start: dar0 = dac0 + dac1 and dar1 = dac2 + dac3 (we have 2 DS boxes in the SAN)
- move all luns to dac1
- rmdev -dl dac0 (gave no error)
- put the DS (dac0) in an other port + change zoning
- cfgmgr on AIX
- troubles
- on 1 server we had 3 dac's: dac1,2,3: no dac0
- on an other server we had 4 dac's: dac1,2,3,4: dac0 was still listed in all_controller for dar0 (together with dac1), so somehow, dac4 did not replaced dac0
-> the solution for these 2 servers was moveing all volumes to dar1, remove dar0, cfgmgr

Some months ago, I had the same problem on 6 lpar's. After 4 weeks of trying _everything_ we reinstalled all lpar's.

So, has someone guidelines for AIX + DS?
I found out that I need a disk on lun 0. I also sometimes have to switch the lun to another controller before I can boot my AIX box.
Any recommondation for other settings for fcs/dar/dac/DS ?


Stef

[moralebr]

Stef,

It's can be SAN-related problem. AIX produces a second WWPN information when SAN switch port is changed.

Basically, dac devices are directly realted to HBA port informations. So, try remove the dac and fcs devices connected through port changed. It can be help.

Regards,
Luciano Martins.

[Stef]

Removing dac's is not working, I tried this before.

And try to remove a fcs device when you boot from the SAN ........

As long as I'm able to remove the dar, it is working. But to be able to remove the dar, I need to deconfigure all the disks so I need a) a second DS or b) boot from internal disks.

Somehow, the dar's are not updating the list of available dac's. It still keeps the old dac somewhere in the configuration.

The next 2 weeks, I will be not at work. I'm going to create a test setup with a spare DS and an lpar to test this. And I'm going to spam IBM untill they find a solution.


Stef