HPUX 11i upgrades

[Adam Noble]

We have a number of servers which we are looking at upgrading in the near future. We have decided to take an approach of cold installing 11i on each server rather than upgrading as we feel its the cleaner approach. One question I have surrounds the non root volume groups. The data resides on EMC disk and I am assuming as long as I export the volume groups, make the neccesary directory structure etc that I should simply be able to re-import them and get visibility of the data. Is this a standard approach to upgrading. The installation and upgrade manual doesn't even touch on it.

[mirco_1]

hi,

the upgrade impact only the root vg (ex. vg00)where there is the system filesystem.

DM.

[RikTytgat]

That should indeed work as you describe. But be sure to use the -m with your vgexport command to have a link between the lvols in the volume group and the name of the lvols (not necessarily in the same order) and save the file on some other server. Also take a copy of /etc/fstab to remember which lvol mounts on which mount point.

After the cold install, use the vgimport command with the -m file you created before.

Regards,
Rik

[Borislav Perkov]

Hi,
You have to install the OS and needed SW only. I think that you are on the right way doing the connections with other vgs. Exporting them, same directory structure and importing the vgs.
Regards,
Borislav

[Alessandro Pilati]

Adam,
before you plan the installation, have a look to these hits useful for your needs:

http://www.docs.hp.com/en/5991-0792/ch03s03.html#d0e2994

http://www.docs.hp.com/en/5991-0792/index.html

Rgds,
Alex

[Gavin Clarke]

Looks like you're on the right track. When we last migrated we did it using Data Protector and restored the data, mind you that's because we changed disk array too.

Yes, I think you should be able to present the disks to another host a bit like serviceguard does.

I would use:

vgexport -s -p -m /tmp/vgNN.map /dev/vgNN
copy mapfile somewhere safe (backup or other host)
vgimport -s -m vgNN.map

Just in case some more info helps people, are you upgrading from 11.00 and which version of 11i are you going to v1 or v2?

[Leif Halvarsson_2 · Accepted solution]

Hi,
Yes, it shoul work as you expect but, to be safe, make sure you have a good backup of all data before starting upgrading.

[Adam Noble]

Thanks for all your help, I just needed to know it was a feasible approach. Just to clarify as there seemed to be confusion, I'm upgrading on the same system not to an alternate host. I have enough info anyway thanks again

[Gavin Clarke]

When I said another host I meant your newly installed system will be just like another host.

Although thinking about this you'll be configuring them to be the same so perhaps my line of thinking was a little confused.

Good Luck with the upgrades.

[D.Blond]

sea attachement, it's simple.

Regards,
Dominique

[Patrick Wallek]

When doing your export of the VG's make sure you use the '-s' option as well so that the VGID gets written in the map file that gets created. You then specify the '-s' with vgimport and you do not have to specify each and every disk that is part of that VG. The vgimport scans all PV's for that VGID and imports all disks it finds.

# vgexport -p -v -s -m /var/tmp/vg??.map vg??

When importing:

# mkdir /dev/vg??
# mknod /dev/vg??/group c 64 0x0?0000
# vgimport -v -s -m /var/tmp/vg??.map vg??

[Bob_Vance]

Note about the "-s" :

If you use the -s option

and you are using distributed allocation

and you have dual HBAs for redundancy

*and* you are alternating primary (and secondary) paths between the HBAs to try to get some data transfer load balancing across the HBAs

(whew a lot of 'if's, but I run into this fairly often),

then the 'vgimport -s' will not keep your nice HBA load balancing scheme.
You will have to manually go back and create the alternating scheme.
However, it's simple to do -- simply vgreduce/vgextend every other primary path.

However, this points out a minor advantage of specifying the PVs directly on the vgimport, assuming that the device names are the same.

However, after a cold install, it's quite possible that the device names could change, in which case the "-s" option is a godsend.


hth
bv

[Mahesh Kumar Malik]

Hi Adam

Your approach seems to be OK. Please ensure there is no issue on the EMC model support on hp-ux 11i and system is patched with latest patch bundles

Regards
Mahesh