SERVICE GUARD needs reboot?

[???_185]

there are two superdome system(11.23pi)managed service guard.and they are 4 packages each other.first server one have ora_pkg1 package is using by shared mode to access data from two servers.
second, server one has pkg1 package is using exclusive mode to access achive and dump files.
third, server one has pkg3 package is waching oracle daemon.
fourth, server one has pkg5 package is checking gigabit card.
server two is like that.
from this weeks, try to upgrade oracle 9.2.0.5 to 9.2.0.6.
does it needs to reboot system? if needs,
i want to know what orders do I have to proceed in SERVICE GUARD.
specifically, for example, first shutdown oracle, second, cmhaltnode -f system1.....

becase this is operating system in permanently, after one system patch is ended,
other system have to proceed work.
Thanks Reguards.

[melvyn burnard]

Unless the newwer version of Oracle requires some network or kernel patches or a kernel tune, I see no reason why you would need a reboot.
Check with Oracle if there are patch or kernel change requirements

[Luk Vandenbussche]

A patch update from oracle, in your case 9.2.0.5 to 9.2.06 doesn't require reboot.
OS patch update are not required in this case

[???_185]

if system must be reboot, what order do I have to proceed?
two system is clustered by mc service guard.
but, it needs to be running for users.
i mean the first server have to be reboot and then second server have to be reboot..
could you check the reboot order?
1. cmhaltnode -f server1 -> in first server
after first server is rebooted,
2. cmrunnode server1
3.cmhaltnode -f server2 -> in second srever
after second server is rebooted?
4.cmrunnode server2
is right? is there anything to add?

[Jon M Zellhoefer]

You won't need to reboot unless you install kernel patches (ie: PHKL_XXXXX). If you end up needing to reboot for some reason that I can't identify, do a cmhaltpkg (pkg_name), then a cmrunpkg -n (node)*the_other_server* (pkg_name), then do a cmhaltnode and reboot.

[Luk Vandenbussche · Accepted solution]

Hi zungwon

1) Stop al the packages on node one
cmhaltpkg packagex
2) Stop the services on node 1
cmhaltnode node1
3) Start package one node2
cmrunpkg -n node2 packageX
4) reboot node1
shutdown -r 0

Perform the same steps if your are updating node one

[Geoff Wild]

First off, I wouldn't do the -f unless I had to...instead, shutdown each package on the node, move them to the other node...with cmrunpkg -n node2 mypackagename...then do the cmhaltnode...

But - if you oracle is on disks that are controlled by MC/SG - then you need to upgrade oracle while the package is running!

Rgds...Geoff

[Hakan Aribas]

reboot steps:

1. shutdown Oracle
2. stop crs (/sbin/init.d/init.crs stop)
3. stop packages (cmhaltpkg )
4. stop nodes (cmhaltnode )
5. shutdown cluster (cmhaltcl -f -v)
6. reboot server (shutdown -ry now)

or

1. shutdown Oracle
2. stop crs (/sbin/init.d/init.crs stop)
3. shutdown cluster (cmhaltcl -f -v)
4. reboot server (shutdown -ry now)

[???_185]

there is ora_pkg pakcages two server is using shared volume.i think it doesn't affect oracle upgrade.
I have other problems.
after this process :
1) Stop al the packages on node one
cmhaltpkg packagex
2) Stop the services on node 1
cmhaltnode node1
3) Start package one node2
cmrunpkg -n node2 packageX
i didn't shutdown node1 system.
and i want to shutdown node2 system in one hour.
and then, i'll upgrade two node system.
how can I shutdown node2 package?
is it OK that just type like "cmhaltnode node2"

last, after reboot two system, if type like this "cmruncl", can these system is being operated normally?"

THANKS REGUARDS.

[Geoff Wild]

Do a cmhaltpkg packageX on node2, then cmhaltnode, then cmhaltcl, then reboot.

When they are up and running - cmruncl

Unless of course you have AUTOSTART_CMCLD=1 in /etc/rc.config.d/cmcluster

then the cluster will start at boot (I don't do this - I always start manually - in case you don't want the cluster up and running - due to a hardware failure ot what have you - IE - you should troubleshoot your system prior to running the cluster in the event of some failure).

Rgds...Geoff