So many answers, so little time.
Overall, most responses said that patch procedure (now called Rolling
Upgrade) works well. You have to use the clu_upgrade command and that the
procedure is well documented and straightforward. It will, however, take
additional time to do the patch than on a non-clustered system.
Responses like these give me more confidence to do this procedure. :-)
Thanks to all that responsed!
Here are excerpts from some responses I thought were worth mentioning....
=====================================================================
From: "Goetzman, Dan" <Dan_Goetzman_at_bmc.com>
> So, I would recommend that you study all the release notes, and form a
> exact procedure as documented for each step in the rolling upgrade. Then
> FOLLOW it exactly. This is no time to "fly by the seat of your pants"
> and "wing it". As the rolling upgrade is a complex process, any
> deviation could cause problems.
=====================================================================
From: Gunther Feuereisen <gunther_at_gfh.com.au>
> It needs to copy a whole bunch of files and make backup copies; so the
> easiest rule of thumb is make sure cluster_root, cluster_usr and
> cluster_var are using < 50% of space. It doesn't actually use it all,
> but this is the easiest measure.
>
> It takes time to copy, so make sure the member you do it from is the CFS
> server for those filesystems.
>
> i.e.
>
> for i in / /usr /var
> do
> cfsmgr -a SERVER=$(hostname -s) $i
> done
From: Alex Gorbachev <alex_at_iss-integration.com>
> Putting a rolling upgrade of PK3 on two 5.1 clustered systems worked
> well, took about 2 hours with coffee breaks :)
=====================================================================
From: Rick Beebe <richard.beebe_at_yale.edu>
> I just did it. It mostly worked and took about 2 hours. I did a rolling
> upgrade so the cluster was never unavailable. Read the docs carefully
> for a rolling upgrade. You basically use /usr/sbin/clu_upgrade to do
> everything.
>
> Pick a member to be the 'lead member'. I chose member 1
>
> clu_upgrade setup 1
> (reboot non-lead members)
> clu_upgrade preinstall
> clu_upgrade install
> dupatch
> (lead reboots)
> clu_upgrade postinstall
>
> Now on non-lead members:
> shutdown now (bring to single user mode)
> init s
> bcheckrc
> lmf reset
> clu_upgrade roll
> (non-lead reboots into patched os)
>
> After all have been rolled:
> clu_upgrade switch
> (reboot all machines one at a time)
>
> clu_upgrade clean
Received on Wed Oct 10 2001 - 20:33:33 NZDT