Changing AIX LUN paths

Hello,

Looking for some guidance to changing how an AIX 6.1 server sees its luns. Presently they're connected via 1GB iscsi but I want to change it to 4GB fiber channel. Anyone done this before and if so is it an easy step?

I'm thinking this can be done in this order:

  1. unmount the filesystems
  2. varyoffvg vg
  3. make the changes on the storage system (EMC)
  4. cfgmgr
  5. varyonvg vg
  6. mount the filesystems

I've never done this before so I'm hoping its that simple but IBM is never that simple.

Any help is much appreciated.

Let me start by saying that I'm not an AIX specialist but I can help in the storage area. I haven't looked at your profile on this forum and you don't state your level of expertise. However, I am aware that EMC have issued very large "connectivity guides" for very many O/S's and this includes AIX.

Obviously we all acknowledge EMC's Copyright of this material but if you search the web for "EMC AIX Connectivity Guide" you can download the book in pdf format. I warn you that it's BIG!

Many years ago when I read the EMC Solaris Connectivity Guide I was parked beside a coffee machine for hours and hours but I learned loads. Perhaps you need to do something similar if you don't already know this stuff. I can't help you any more than that. AIX specialists on here should be able to help you more.

Good luck. Hope that helps.

Thanks for your reply. I tried the menthod I listed and that didn't work. Turns out, AIX set the paths to each disk, since the disk is going over iSCSI, the disk attributes has the iSCSI target name in the connection so it didn't work.

I'll seek out the EMC connectivity guide and give that a go.

Thanks.

If anyone else has done this, I'd be interested in hearing.

The LUN that AIX recognises will be based on the serial number it is presented with. I no longer have AIX available, but I think you can see it in the Z00 or Z01 line in lscfg -vl hdisk7 or whatever.

You should be able to add the new links in from the SAN and get wider output from lsvpcfg

If you can, then you can delete the old hdisks and remove the old SAN definitions. This way, you don't need to take them offline.
Obviously, don't start removing things until you are certian that the new hdisks (discovered with cfgmgr -S ) are merged into the existing vpaths.

Make sure that you remove the old hdisks before you try to remove the LUN at the SAN end or you will get lots of hardware failure messages.

I hope that this helps.
Robin

I use iSCSI with my VIOS. I then export those disks using vSCSI to the partitions.

As it is iSCSI - it is dependent on how your IP network connects.

iscsi0    192.168.90.53    3260 iqn.2004-04.com.qnap:ts-659:iscsi.x061.c806e6

My guess is that if you can change the IP network address from the 1GB network to the 4GB interface - just change those.

If that is a potential solution I, or someone else, can tell you how that can be done without taking anything "offline".

And, if you want to do it offline - it is not just taking all the filesystems offline (aka unmount) but also varyoffvg of the volume group, and then rmdev -l hdiskX (or rmdev -dl hdiskX ).

And yes, if you varyoffvg the volume group, you will probably use cfgmgr and varyonvg , and mount ... to bring everything online again.

So - what does the IP network look like?

My single 'link' to my NAS gives my 13 iscsi luns. An older, simpler, device had one lun per iscsi entry.

So, how the luns look|appear depend more on the device than on AIX - imho!

Hope this helps!

I don't know this side of EMC, but let's assume, that nothing will be changed on the underlying device during changing the connection type, then you need several additional steps:

-1. make backup
0. notice all PVIDs of all disks, you have
2.5. exportvg all volume groups, rmdev all the disks, you plan to change
4.5. make sure, that you see the same PVIDs on the new disks. If you don't - go back and doesn't try to do anything further.
5. importvg all the volume groups, you exported before.

It could be also good to notice VG major numbers. Sometimes they are useful.