Upgrading to HPUX 11i from 10.20 space questions

[samd_1]

A few questions concerning disk space and cold install. I have included a Word doc that is a screenshot of my disk allocation as of right now. As you can see from it the root volume is 140 meg total with 94 meg free. /stand is 87 meg total with 59 meg free. /usr and /opt have close to 600 meg and /tmp has close to 200 meg free. I'm not clear based on the install guide of a few things.

1. What is the disk space needed for the cold install? From what I see it only mentions the need for 4 Gig in the root. Everything else seems to be for the update method.

2. If root does need 4 Gig does that mean I need another drive in my current configuration? Or should I/can I reconfigure my volumes and take space from another volume to give to root? IF so then I take I need to make a backup and restore the filesystem I'm taking space from because this will be deleted.

3. Does the cold install wipe out your whole volume group or just the root filesystem and other system related filesystems such as /opt, /stand, /usr?

[Pete Randall]

Sam,

It's going to depend on your environment: 700 or 800? what 11i operating environment you'll be install? what applications you'll be loading?

As far as filesystems are concerned, a cold install will replace everything in vg00, typically /tmp, /var, /usr, /opt, /stand, and possibly /home.

Hope this helps,
Pete

[Sandip Ghosh]

I don't know the size of the root disk. But as a standard I would have kept like this

/ - 140 MB
swap - 2 GB
/stand - 80 MB
/opt - 750 MB
/usr - 750 MB
/var - 750 MB
/tmp - 500 MB
/home - as per requirement.

When you do the cold install all the partitions on vg00 will be effected.

Good Luck!

Sandip

[Patrick Wallek]

If you do a Cold Install of HP-UX 11i, you WILL LOSE everything that exists on your VG00 disk(s). The cold install completely wipes out the current VG that is on the disk.

I would definitely recommend a 4GB disk as the absolute minimum, a 9GB would be better. You would have some breathing room with a 9GB, you won't have much with a 4GB.

I did a 11i install the other day and I made / and /stand 256MB (generous, I know) /var and /usr 1GB and /opt 2GB, I think. /tmp and /home are entirely up to you.

[S.K. Chan]

First of all before you do anything, I would recommend you do some "housekeeping" of your vg00. Typically vg00 should not have "user data" in it but in you case it looks like you got ..
lvol9->lvol13
in vg00 which in my opinion needs to be "taken" out from vg00 so that vg00 ONLY has bootable and system files. You probably got more than 1 disks in vg00 ?
# vgdisplay -v vg00
If so, then you should vgreduce it so that you can free up "some" disks and create a separate VG (vg02), migrate files from vg00 (lvol9->lvol13) to this new VG.
Once you got vg00 clean up it is easier to migrate to 11i. Cold install will wipe out vg00 which is okay since you got all your data separated in a different VG. You would then use vgexport/vgimport to bring vg01 and vg02 back onto the newly installed system.

[James R. Ferguson]

Hi Sam:

First, don't upgrade, do a cold-install. Upgrades from 10.20 to 11.x have a much higher rate of failure and/or problems than cold-installs.

As Patrick noted, 4GB disk will work (barely) but at least 9GB is much more comfortable.

I always look to isolating my vg00 on one physical disk and mirror that. I see from your post that you have lots of application filesystems on vg00. I would *not* do this from a performance, a backup and a recovery standpoint. I'd establish separate volume groups for this purpose. The sizes of your logical volumes that represent the OS filesystems are good. Here too, be generous, especially if you have 9GB or larger disks. The '/var' filesystem is a filesystem that can consume considerable space, especially as patches are applied. '/var' also grows larger if your server is actiing as an Ignite server holding Ignite archives. Consider that too.

I am very generous in sizing it as well as 'tmp'. '/tmp' deserves more space than sometime afforded becuase this is the standard staging point for unsharing patches and building patch depots. There are workarounds during patch application but in my opinion a well-sized /tmp filesystem makes life easy.

The root filesystem is generally misused and hence may tend to grow in size. If you want to catalog lots of files that have no other "home", create a mountpoint and appropriate directories somewhere else.

'/usr', and especially '/opt' are dependent on what 'applications' you add -- for instance 'ansiC', 'glance', etc.

While you can certainly extend all of the standard vg00 filesystems (or re-Ignite later) the cold-installation process is an excellent time to reassess what you run and what you plan to run, and correct some of the annoying bottlenecks you have had.

Regards!

...JRF...

[samd_1]

This is a K580 B.10.20 9000/800. Oracle is the only real application that will installed and it is a development environment. Therefore we like to keep the datafiles and other Oracle files in the same location as the production environment as best we can. I have included the results of the "vgdisplay -v vg00" command that S.K. Chan suggested. Therefore does this mean I need another disk? Or from what you guys see is it possible to rearrange this disk space so my root can have 4Gig and the other filesystems their required spaces?

[Pete Randall]

Sam,

I'd have to say you'd be better off with another disk. They're relatively cheap nowadays, it will ensure that you have plenty of room, and it will allow you to separate your user data into a separate volume group on a separate physical device. That way if you lose your root disk, you can rebuild it without having to restore all your user data.

Good luck,
Pete

[Bill Hassell]

The root (physical) disk size should be 4Gb but the / directory should be no more than 150-200 megs. The other size recommendations for the opsystem are fine.

However, from your bdf listing, it appears that you have intermingled the Oracle files with data stored on vg01 and backup+peps things on vg00. To keep your sanity, it is best to keep opsystem things on vg00 (only) and everything else on a separate volume group.

Now when you cold install, it will not touch the vg01 disk(s) at all. So now would be a good time to migrate non-HP-UX directories out of vg00 and onto a new disk, either part of vg01 or maybe create a vg02. Note that Oracle only cares about the mountpoint, not the physical disks so everything will work OK once the data is moved and the local vg00 directories are used as mountpoints.

Then before you actually cold install, download the latest Ignite/UX package for 10.20, run: make_tape_recovery -AIv and store this away for safekeeping. Do your cold install and override the lvol sizes to match your current values. Then apply the Support Plus patches and finally, you can import the volume group(s) using SAM. Once imported, the mountpoints can be added back to /etc/fstab and you'll have your data online.

[samd_1]

Yeah but can't I do the following? I do not know the exact commands to do all this, but I'm wondering if it's possible.

1. Take away some of the disk space from the other VG's. Make sure this is enough to hold the user and app data in VG00 with some room to grow.

2. Create a vg02.

3. Move the use and app data volumes into vg02.

4. Now with the leftover space allocate accordingly to the vg00 volumes so there is enough space for the cold install.


I keep looking in the manual and all I see for the requirements for a cold install is a 4 Gig root disk. Is this correct? I see what some of you guys recommend for the volumes but since I don't have gobs of space I'd like to be somewhat accurate in space I allocate. Thanks guys and gals.

[Jim Butler]

Sam,

If I were you, I would upgrade. There are a couple of issues -
One - What version of oracle are you running now, and are you going to upgrade oracle also as part of the deal? As part of this thought, if your version of oracle supports 11i, you may want to update it after the upgrade. If it doesn't support 11i, you may want to upgrade oracle first, then upgrade.
Two - Note how your vg00 is over 30 Gb - Is it made up of one disk, or did you combine multiple disks to this vg?
Three- You may want to consider an export of your oracle data, and then do a cold install. You will need to do this if you want to do a cold install, since you have the vg00 problem (i.e., OS and Apps are on VG00)
After the upgrade, run cleanup to remove the references to the 10.20 OS - the suite of swlist, swinstall, swmodify, swverify, etc tools help with the post migration issues. I would recommend running cleanup -c 1 to clear out old 10.20 patches, etc.
cleanup -s when you are done will also help.
good luc

[MANOJ SRIVASTAVA]

Hi Sam


Here are some of the guideline you may like to look at :

http://docs.hp.com/hpux/pdf/5185-6593.pdf

check for the disk spcae and memory requierments , these are the minumum so if you aloading softwares app etc then you need to bump them

It is alwasy ggo as you suggested that move the data acroos to vg02 and have some more space in the vg00 . Also please keep some buffers bcoz there are filessytems which may need to ne increased on demand.


Manoj Srivastava

[samd_1]

OK I've looked at the PDF before and that's what is unclear.

Anyway are you saying that is is ok to create a vg02?

IS it ok to just take space away from existing volume groups?

[MANOJ SRIVASTAVA]

Hi Sam

Ofcourse the proper way to do that is to restructure your vg00 , it has got 4340 PE each of the size 8.00 Gb ,ie the size is 32 GB , it give me a impression that you are bigger disk size for the root disk . Incase you are at hte pallning stage then I woul suggest is if possible take a sammler size disk for the root disk like 8.0 GB and use this one for the Data purposes only . If not then you the question come is what will you sacrifice in case you ahve 2 volume groups on the same physical disk ( ahveing one controller ) to access root area , swap and and also the data , the better idea in case you dont have any options then have the structure like it is .


Manoj Srivastava

[samd_1]

My current config is 3 32 Gig disks in the server.

vg00 uses the first disk
vg01 has 4 logical volumes with 2 located on each of the remaining disk. (that's right there are 64 gig total allocated)

What I was trying to find yesterday is the advantage to having more than one volume group. I see there are mirroring reasons mainly. Also there are organizational reasons. Another thing I'm not clear of (although the last response seems to indicate I can do it) is whether I can create 2 volume groups on one physical disk. For instance ideally I would like to vg00 to only take up part of the first 32 Gig disk. This was I can allocate the rest for another volume group. Also if this is done will the cold install wipe out the volume group or the physical disk.


Since this is a development server I think what I need to do is not worry about the performance issues (separating indexes to a different drive from the tables in Oracle). This was I can do this reorginization and use the disk space fully without buying more disks.
Not the next question is expanding the root, /stand, and the swap since they are all contiguous. fun.

[Pete Randall]

Unless I'm sadly mistaken, Volume Groups are made up of whole physical disks (man pvcreate and man vgcreate). You can't split a physical disk between two different volume groups.

What everyone seems to be suggesting is to isolate your root volume group. The best way in your case would be to get a new disk (4GB minimum, 8GB preferred) and use this for your root volume group, separating all other data into different volume groups on different physical disks.

Good luck,
Pete

[samd_1]

Looks like I need to do some reading because one person says I can put 2 volume groups on one disk. The other says no. Hmmm.

I still don't get the point of volume groups other than for mirroring or performance reasons.

[S.K. Chan · Accepted solution]

Sam, let me try to clear things up for you, sorry if we are causing more confusion to you than helping you to understand.
Ok this is your situation..
-3x32GB
-vg00 has 1x32GB (diskA)
-vg01 has 2x32GB (diskB and diskB)
-no mirroring
The problem is you got a 32GB on a root VG (vg00) which is a "wastage" since the "right" strategy is to only keep OS fileystem on vg00, not data FS for reason of disaster recovery, hence a smaller (eg: 4GB) root disk is recomended.
Hence you want to allocate the "un-used" space in diskA (vg00) to vg01. Well you CANNOT do that, you can't have a disk belonging to 2 different VG (vg00 and vg01).
It would be "ideal" if you can move the "data" LVs from vg00 to vg01 (depends on whether you have enough space in vg01 or not). That way the only LVs left in vg00 is for OS.
If you do not want to do that then it's fine (since it's a development machine) but make sure you got a good backup.

[samd_1]

Thanks S.K., I won't mind wasting some space. Since this is development there is so much disk space right now anyway. Most likely I'll keep vg01 as it is and just move the volumes in vg00 over there.

I'll need to look these up in the HP books.

[James R. Ferguson]

Hi Sam:

To help you better understand LVM terminology and concepts, have a look at:

(1) the man pages -- 'man 7 lvm'. This gives a quick overview of concepts and the commands used to manipulate LVM structures.

(2) The "HP-UX 10.0 Logical Volume Manager White Paper", Technical Knowledge Base #OALWP03950320. While this is an "old" document, it remains quite informative.

(3) The "Managing Systems and Workgroups: A Guide for HP-UX System Administrators" manual's chapter on "Administering a System: Managing Disks and Files":

http://docs.hp.com/hpux/onlinedocs/B2355-90742/B2355-90742.html

You might want to read these in the order noted.

Regards!

...JRF...

[samd_1]

Thanks for the help guys.