Upgrade to HP-UX 11.0 and internal disk replacement

[Jeff Noblitt]

Hi.

We are trying to plan an upgrade from HP-UX 10.20 to 11.0 in order to upgrade to a newer version of Sybase. Current root disk/file system is on a 2GB Raid 1+0 array. HP O/S upgrade assessment recommended 9GB for O/S (K450, 2GB RAM), so we have purchased 2 18GB disks to replace the 2 current internal drives. Apparently, if these 2 18GB disks are installed configured as the Raid, all 18GB will be mirrored via raid, so a lot of recently purchased disk space will be basically unavailable. We need the space.

I would like to install in such a way that only 5GB on each new disk is mirrored (for the O/S), 4GB on one disk is primary swap, and the remaining 22GB is available as unmirrored applications space. Will MirrorDisk/UX accomplish this?

Also, its been suggested that the only way to replace the current internal disks with the new, larger ones is to wipe out the entire system and start from scratch with a cold install of 11.0 . Since it is a production DSS, I would like to avoid that if possible.
Could MirrorDisk , or some other method, be used to mirror/copy the O/S to an external array, then boot from the external array as an alternate boot path, then remirror the O/S back to the new internal disks?

Thanks, Jeff.

[Manju Kampli]

Jeff,

If you can install these two 18GB disks as stand alone disks then you can use mirror/ux to do mirror for your root file systems where system files sits ( /, /stand, /usr, /var, /opt etc) and mirror these file systems to second disk. The rest of the space can be used for your application with out mirroring. This will save you the disk space and mean while if one disk fails the system will run but your application data will be lost since there is no protection.

I am not sure whether your array encloser have a empty slot to fit these two new disks. If it yes. then you can leave the OLD disks untouched and install 11.0 with the new disks and if you want to go back to 10.20 you can boot from the OLD disks.
Hope this helps,
Manju

[Michael Tully]

Hi Jeff,

The advise given to you should work,
There is a few things that I would
like to add, and that is to make sure
that you have at least one good
make_recovery tape at hand to restore
necessary system files onto your HP-UX
11 system from your 10.20 system. The
other piece of advise that I would like
to offer is that you should also have
an exported copy of you lvm at hand
also so you can import any volume
groups that you wish to see on the HP-UX
11 system. Make sure that you only use
the -p option.

HTH
Michael

[Magdi KAMAL]

Hi Jeff,

Of course, you may mirror only the file systems( logical volumes ) using by unix and leave other logical volumes without mirroring.

The mirroring operation is not an automatic task, but you need to map and execute such commands explictly.

example :

If you have your logical volumes lvol4,lvol5,lvol6 on disk /dev/dsk/c1t2d0 and you want to mirror only lvol4 and lvol5 on disk /dev/dsk/c2t2d0, you may issue the following commands :

pvcreate -f -B /dev/rdsk/c2t2d0
vgextend /dev/vg00 /dev/dsk/c2t2d0
mkboot /dev/rdsk/c2t2d0
mkboot -a "hpux(;0)/stand/vmunix" /dev/rdsk/c2t2d0
lvextend -m 1 -A n /dev/vg00/lvol4 /dev/dsk/c2t2d0
lvextend -m 1 -A n /dev/vg00/lvol5 /dev/dsk/c2t2d0

and so on for other hp-ux logical volumes.

hope this helps.

Magdi

[Magdi KAMAL]

Hi again,

it's recommanded by hp to fresh install hp-ux 11.0 and not upgrading ! I beleave this could avoid future troubles and headeck.

I had done it ( fresh installing hp-ux 11.0 )for my costumer and it works fine, using the Ignite_UX.

You may do the following steps :

1. Backup your entire actuel vg00 ( hp-ux 10.20 ) with the make_recovery command, this create the lif area on the tape and follows it by all vg00.
2. Install fresh HP-UX 11.0 on test system, and all its needing applications and software.
3. When all problems are solved, perform another make_recovery to save actuel installation.
4. Restaure this last backup ( Ignite_UX ) on production server by :
- Booting the system.
- Interrupting the boot process in ISL level.
- Boot from alternate path which is tape .
- Interrupt the download process from tape and give the production hostname and IP address, ...
- follow instructions and installation is OK

Conclusion :
You will do the installation of hp-ux 11.0 and the make_recovery only once and will use that tape n times to bild what ever server it is.

Notice :
You will need one make_recovery by server type for hardware architecture specifics ( example : One for N class server, One for L, ...)

Good luck

[James R. Ferguson]

Hi Jeff:

MirrorDisk/UX mirrors *logical volumes*, not disks. I don't see any reason not to mirror all of the logical volumes of vg00. You can't assign multiple volume groups to the same physical disk(s) either, so, in the case of a root disk, the larger physical disks sometimes see some "waste" of space. The smaller capacity disks are harder to find, and have less usefulness. In the case of your root volume, build some well-size swapspace on it, and size your logical volumes in a fashion that allows plenty of flexability.

I strongly recommend a cold-install if you are moving from 10.20 to 11.x. This is the least error-prone method. You have the ability to size your logical volumes according to your needs. There are a multitude of threads here with suggestions for a successful installation. Since you have 2GB root disks, I don't see any good choice but a fresh install onto larger physical disk.

Consider readdressing your existing internal drive to a lower SCSI number and addressing your new drive at the more optimum high value. Cold-install to your new drive leaving your old drive as a safety-net.

...JRF...

[Jeff Noblitt]

Thank you all for the excellent advise. I'm not a Unix SA, so I am trying to piece this all together so that (hopefully) our upgrade goes smoothly.
It's unclear if our GSC built-in fast/wide SCSI has 2 available slots, but I like the idea of installing the 2 new 18GB drives as stand-alone (rather than RAID 0+1), and using mirror-us (which we don't currently have) to accomplish the O/S mirroring. If we had that arrangement, and mirrored 5GB (/stand, /var, /usr, /tmp, /opt, /home) from disk 1 to disk 2, would that restrict both disks to only VG00? The comment about not being able to assign multiple volume groups to a physical disk has me a bit confused. Would I be unable to utilize the extra space as part of a different volume group?

For make_recovery, the man page says it works to boot from DLT for K-class systems that use the HSC bus. I'm not sure what the bus type is (GSC ?).

A lot of good information. Thanks, Jeff.

[Scott Riley]

Jeff,

A physical disk may belong to one and only one volume group, so using your two new disks for vg00 will resrict all volumes on those disks to the vg00 volume group. So yes, you would be unable to utilize the extra space as part of another volume group. However, there is nothing preventing you from creating new logical volumes within that extra space and mounting filesystems on them. The only limitation is that these new logical volumes must stay within vg00.

Just be careful about what you put on those drives, you could impact overall performance.

> I'm not sure what the bus type is (GSC ?).

GSC implies HSC.

- Scott