1) I agree with hellonair - that it is not a good idea to be using shared storage for rootvg maintenance.
2) iirc (if I recall correctly) we advise to bring a nods being updated into a dawn status - move any active resources to the remaining node(s), stop hacmp, and start updates.

Before starting any updates on a HACMP node you need to verify that your application/service is supported in the new environment. We have situations discussed here where applications did not work after an update.

With HACMP (PowerHA) I strongly recommend doing an alt_disk_install plus update, so that should the application and update not work as expected you just need to reboot the node to restore the old situation.

Re: 3G in rootvg left. I guess I wonder how large your rootvg disks are - assumming they are mirrored. For an update, that should be plenty - as temporary storage. I generally only have 1G set aside for an update. However, I am wondering a bit about what else you have in your rootvg - if it is filling up. Applications and data should not be in rootvg, but in seperate volume groups.

3GB of free space shouldn't be a problem surely for any update - this isn't Windows Bloatware   
You could run three complete AIX web servers and the entire rootvg site in that space.

If you are updating Oracle, that and any other applications, should always be in at least a separate filesystem or two,
and ideally in a separate volume group.  You can add another volume to a volume group to give it more space and remove it later.  Increasing filesystems is easy, shrinking them back again may not be ! 

Sounds like you have isolated Oracle and I believe your passive node updating is the correct approach.