Space constraints on vg00

[rmueller58]

We are preparing for an upgrade to the OS from 11.00 to 11.11 (possibly 11.23) and Informix 10.

We have an N4000 with 2 mirrored 9gb drives for VG00, we have 2 SC10 subsystems in RAID0 with loads of free space.
We are also in process of purchasing a HP SAN. We will have a fibre channel connection to it..

our VG00 is maxed, and according to the install docs we don't have enough free space in /opt /usr to do the OS upgrade.
We are considering moving /opt to a different VG with enough allocated to handle what ever might get thrown at us, and recapturing the space from VG00 previously used for /opt and resize /usr

We are looking for methods and recommendations on how best to make this work. Any thoughts, suggestions or recommendations would be appreciated.

Rex Mueller - Educational Service Unit #3
Omaha, NE

[James R. Ferguson · Accepted solution]

Hi Rex:

I would take a different approach. I'd purchase two new drives for vg00 (e.g. 72GB) and at the point of upgrade, swap out your existing drives for the new ones and *cold-install* your operating system.

This gives you (both a fallback/backup of your old operating system, *and* the ability to cold-install which is far cleaner than an upgrade.

Too, if you are thinking of transitioning from 11.0 to 11.23, you must step through 11.11 unless you cold-install.

Regards!

...JRF...

[Pete Randall]

Rex,

One thought would be to combine the two 9GB drives and then put the mirror on the SAN or the SC10, or just move the whole VG00 onto external storage.


Pete

[Torsten.]

Hi,

do not move+link /opt or /usr or any other part of vg00 to another VG - this will bring you in deep trouble some time later.

Follow the suggestions and replace your disks with larger ones or use a drive from your SC10 JBOD.

[rmueller58]

I concur with James about replacing the 9gb drives. I got a quote on a pair of 36gb drives for under $500. It would be the best solution for long term reliability. The problem becomes the outage time requirements.

1. Ignite existing system
2. Split mirror
3. Place one drive ignite the drive
4. Boot to new drive
5. Put 2nd drive in
6. restore mirror
7. Adjust file system size.

Any estimates on time required to do this based on a 9gb to 36 gb migration?

[Torsten.]

You can adjust the LVOL size while restoring the ignite backup.

The speed depends on your tape drive.

I took a backup using a fiber channel connected LTO3 drive in less than 10 min.

[rmueller58]

Torsten

We have a DDS3 and a DLT.

[rmueller58]

I have to outline options, assuming the disk purchase gets shot down..

If OPT is sheerly add-on packages would there be any harm in moving it,

say from VG04 (which I have 10gb free on.)

create a new FS on VG04 for /opt?

My big concerns of the items below are the raid4si, perl5, hpnpl, ignite, hparray.

I am not talking about splitting usr or opt just moving one to another VG.

I personally think pursuing the move is the labrabor chasing it's tail but I know this is what management is pushing for. They are spending 200k on a new SAN but god forbid I get $500 for a pair of new disks on a mission critical system..

[Pete Randall]

Rex,

One last thought:

I think the N-class has more than just two internal slots. Assuming that's true, you could get an 18GB drive quite cheaply, combine the two 9GB drives and mirror onto the 18GB drive.

Good luck,


Pete

[rmueller58]

Thanks Pete,

I will check into that. If true that would sure save a lot of sweat blood and tears.

[James R. Ferguson]

Hi (again) Rex:

Beware of trying to simply use 'vgrestore' to enable a larger replacement drive. Your LVM geometry (in particular, 'max_pe' and 'pe_size') was set during the original 'vgcreate' of vg00. This will be replicated during the 'vgrestore' and you may find that you simply don't have the ability to use ALL of the physical extents that comprise your new drive.

As Torsten said, Ignite allows you to rebuild vg00 (beginning with a new 'vgcreate') AND resize the logical volumes. Simply choose the Advanced Installation mode to enable this.

Your management needs to understand that keeping old 9GB drives for the sake of a few hundred dollars compromises and complicates their goal to move to a supported operating system (11.0 isn't anymore) on a mission critical server.

Regards!

...JRF...

[rmueller58]

Thanks James. edu is feast or famine. right now famine because of other dept expenditures. But $500 is a drop in the proverbial bucket. we spend that much on toilet paper in a day. :D

[Steven E. Protter]

Shalom Rex,

If you can't get disk , you an pull one of the mirror drives out and preserve the system for reference and recovery if need be.

The most important issue is that cold install is the only reliable way to go.

JRF's disk plan is a best practice, but even if vg00 isn't mirrored you can do an Ignite make_tape_recovery backup prior to starting and have a good backout plan.

You really need to do your homework though.

Have the necessary patches for Os requires by informix, hardware patches, firmware, the whole plan scripted. Then all you have to think about is the surprises that come up.

SEP

[Marcel Burggraeve]

The list of 7 steps to take can be shorter which will save time.
Igniting 9GB with the tapedrives mentioned should be doable in 60-90 minutes.
Restore should take about the same time.
So in total it could be done in about 3 hours.
If you really need to have as little downtime as possible it would only take about 90 ( including disc removal and placing new ones ) minutes since the ignite tape can be made during production.

Step 2 is not needed if you replace both drives with 36GB ones.
Just mark which one is on top ( or bottom ) and remove them both leaving the mirror intact.
Step 3 and 5 can be combined.
Since you can remove both 9GB drives at once you can also place the two new disks at the same time which saves some time in putting in the new disk.
Step 6 can be done when the system is in production again and no downtime is needed for it.
Step 7 can be done during the ignite restore if you interrupt the process and go with an interactive install.

If really needed a minimal downtime of something like 90 mins should be possible and if you replace the bootdisks with bigger ones you can always go back to the old situation in a couple of minutes if something goes bad.

[Marcel Burggraeve]

Totally forgot about the HP-UX upgrade in my previous answer.
If time really is such an issue don't even think about upgrading but just remove the old disks leaving the mirror intact and do a cold install of the newer HP-UX version.
There's so many things which could go wrong during an upgrade making you loose way more time then you can afford.
Loads of discussions can be found on this forum regarding succeeded and failed upgrades.
Unless you have loads of time available to repair a failed upgrade ( since you're on a tight schedule, be aware of Murphy ;-) you're most likely much better of with a cold install.

[rmueller58]

Thanks Marcell

[rmueller58]

We ignited the original drives, removed, and restored to new pair of larger drives..

Thanks all