We all know that nimadm and multibos don’t play well together (at least that’s the case right now). If you're planning on migrating a system to a newer version of AIX (6.1 to 7.1 for example) you must ensure that your BOS LVs are reset to their traditional names and that all standby BOS instances have been removed. It’s clear that if you plan on using nimadm, to migrate a NIM client to a new version of AIX, you should first remove any and all instances/remnants of multibos, on the client system, before you start.


What’s not immediately clear is that even the NIM master should be cleansed of all evidence of multibos activity.


For example, let’s say you use multibos to install a new TL and SP on your NIM master e.g. AIX 7.1 TL1 SP4 to AIX 7.1 TL1 SP5. You reboot the system and run from the standby BOS instance (leaving the original instance in rootvg, just in case you need to back out).


You then attempt to migrate a NIM client, using nimadm, to a new version of AIX. The operation fails at phase 8, with the following error message:



Executing nimadm phase 8.                                              


Creating client boot image.

ls: 0653-341 The file /dev/hd5 does not exist.                         


At first you might think that this error relates to the NIM client. You check the client system…..there’s absolutely no evidence of multibos on the system. Good.


The issue lies with the fact that the NIM master is enabled for multibos and running from the standby instance.  Restarting the NIM master, on the original instance, booting from hd5, rather than bos_hd5 and then removing the standby instance (multibos –R), will resolve the issue. Now, nimadm will run without error.


AIX Updates With Multibos



Working With Multibos



Migrating to AIX 7.1 with nimadm



Migrating to AIX 6.1 with nimadm