topic Re: Need a refresher in Operating System - HP-UX

Need a refresher

Michael Elleby III_1 — Tue, 19 Mar 2002 23:30:27 GMT

I have a node in an SG cluster that will need to be brought down for an Oracle upgrade. However, the package that it runs may need to have certain volume groups remounted for use. Here are the steps I'm considering

Halt the package.
Halt the node.
Perform a vgchange -c n vgname
Perform a vgchange -a y vgname
remount file sys on logical vol

After the work is done:

umount filesys
vgchange -a n vgname
vgchange -c y vgname (since the cluster will still be running)
restart the node

Any tips or things I might have missed?

Re: Need a refresher

A. Clay Stephenson — Tue, 19 Mar 2002 23:46:27 GMT

Hi:

I would simply halt the package then do a vgchange -a e. The vgchange -c n is not needed
as long as you have not halted the node.

Re: Need a refresher

Michael Elleby III_1 — Tue, 19 Mar 2002 23:50:02 GMT

So, you're saying that I don't have to remove the 'cluster mark' on the vg???? I thought you had to before you could reactivate the vg in order to access it outside of the running cluster.

I do understand when I halt the package, it would deactivate the vg, I just wanted to be sure...

Mike-

Re: Need a refresher

A. Clay Stephenson — Tue, 19 Mar 2002 23:56:23 GMT

Hi again:

I am saying, leave the node as part of a running cluster. Don't do the cmhaltnode. Your box can then act as an adoptive node for other failed packages. If for some reason, you actually need to shutdown that node then your first method is correct but if all you are going to do is upgrade Oracle then then is no need to halt the node but simply halt the package. By the way, don't forget to re-enable package switching when you are finished.

Regards, Clay

Re: Need a refresher

Tania Ap. Serpa Martins — Mon, 25 Mar 2002 18:39:00 GMT

Dear Michel,

I used the same way that you describe, but we wrote 2 scripts to help us to mount and umount the filesystem. First you need to halt you package:

- cmhaltpkg

- to mount your filesystems you can use the script below:

vgchange -c n /dev/
vgchange -a y /dev/
# Check the filesystem that you will mount
fsck -F vxfs /dev//rlvol1
# Mount the filesystem
mount /dev//lvol1 /home
# Configure Cluster's IP Address if you need
ifconfig lan0:1 151.87.20.10 up

- To umount your filesystems that you mount with the script below:

# Unconfigure Cluster's IP Address that you
# configured before
ifconfig lan0:1 151.87.20.10 down
# Umount the Filesystem
umount /home
# Activate Cluster in VG
vgchange -a n /dev/
vgchange -c y /dev/

- Start the package:
cmrunpkg

In the case above, I consider that MC Service Guard daemons are running.

Bye,

T??nia

Re: Need a refresher

Michael Elleby III_1 — Mon, 25 Mar 2002 20:26:21 GMT

Hello Tania-

We wound up not having to mount any of the filesystems that were activated in the packages, since the only fs's being mounted were the actual database files and not the files in ORAHOME (which I found out was in vg00).. Only had to nfs mount some filesystems from remote box that contained the files that the dba required.. So actually, just halting and restarting the packages was all I needed to do..

Thanx for your input..

Mike-