Hello all!
Thank you, colleagues, for your suggestions. I have done the upgrade of SF HA
5.1 to RP2P2 HF3 as well as Solaris Update 9 to Update 10 last Sunday on
four domains with more than 40 zones in total. This is was not quite simple
but it is possible . I can give you my option of doing it but first of all
just a few notes...
I. No Solaris Live Upgrade.
1. Local zone root on shared storage on veritas volume (vxfs)
How to upgrade? Zone must be in installed or running state to get upgraded.
But installrp cannot run if it "sees" mounted veritas volume.
2. Local root is on zfs. Local Zone root on shared storage variant using
VXFS as its local file . ZFS is under DMP control (dmp native support is on)
Yes, this is not difficult to upgrade SFHA only. There are no many patches in
the bundle. But if you must upgrade Solaris version? But if you have 16 local
zones for example?
it takes time to upgrade them all. Sure , you can perform the upgrade
process in parallel (/etc/patch/pdo.conf num_proc =16). But if you get any
problem with new patches (getting some new bugs) or if you get any error
during upgrade (and this is quite possible) you won't be able to roll back
quickly. This is because you must upgrade all nodes of your cluster after
upgrading local zones on one node.
Summary.
1. There is no phase upgrade in that case. This can be unsafe.
2. Installrp does not work if it sees dmp native support switched on.
(with 5.1. SP1 RP1/RP2)
3. It can take much time if you have many zones and many patches.
4. If you have six nodes in cluster you must switch all of your
tasks to a single node. This is not always possible because of lack of
resources (cpu, memory etc)
3. Zone root on local storage using VXFS
If you have six nodes you must have 6 downtimes! This is unreal
plan! This is not suitable.
plus 1-2-3-4 from above.
The only practical way to upgrade the cluster is to use Solaris Live
Upgrade. To be continued in part II..
Regards, Pavel.
From: Anish Vaidya [mailto:anish_vai...@symantec.com]
Sent: Saturday, March 31, 2012 3:10 AM
To: Anish Vaidya; Цветков Павел Анатольевич; Venkata Reddy Chappavarapu;
mamatec...@mamatec.com <mailto:mamatec...@mamatec.com>
Cc: veritas-ha@mailman.eng.auburn.edu
<mailto:veritas-ha@mailman.eng.auburn.edu>
Subject: RE: [Veritas-ha] upgrade VCS 5.1 with many zones
Pavel,
We strongly encourage using CPI installer for upgrading the stack in an SFHA
cluster involving zones. Anyway, if you want to know the manual procedure, here
is the high level procedure.
1. Zone root on shared storage using VXFS
You need to patch the zone only once.
a. Move all service groups containing Zone resource to one system
(preferably).
b. Bring down all the APP resources on top of the Zone. Keep the Zone
running.
c. Stop VCS (hastop -local -force). Upgrade the SFHA packages. At this
time, SFHA packages are upgraded in the global and local zones. You may have to
reboot this node as the VXFS driver loaded in kernel is still of the lower
version.
d. Upgrade the other node in the cluster after stopping VCS (using "hastop
-local -force"). Upgrade the SFHA packages. Please note, no zone is
installed/running on this node.
e. Start VCS on 2nd node. Bring up the relevant Zone service groups.
2. Zone root on local storage using VXFS
You need to perform the procedure similar to 1a, 1b and 1c on all nodes in the
cluster. This is because, each individual zone needs to be patched on all nodes
in the cluster.
Hope this helps.
Regards,
Anish
_______________________________________________
Veritas-ha maillist - Veritas-ha@mailman.eng.auburn.edu
http://mailman.eng.auburn.edu/mailman/listinfo/veritas-ha
