HP-UX 11i disk space

[Stuart Powell]

We are planning to move our ERP system to model 7410 servers in a two node ServiceGuard cluster. One of the big questions is how big of internal drive should we purchase. We will have at least 12 Gb of RAM so we'll need at least 24 Gb of swap space(per HP). We will also have internal mirroring of the boot disk. Database and applications will be on a SAN. Any input from experience would be greatly appreciated.

[Rodney Hills]

See Bill Hassel reply on swap space calculation-

http://forums.itrc.hp.com/cm/QuestionAnswer/1,,0x9d7b42308663d611abdb0090277a778c,00.html

You don't have to allocate as much disk space as you think.

My 2 cents...

-- Rod Hills

[A. Clay Stephenson]

The old rules about 2-3x swap to memory very seldom apply. Note that this would rise to 4-6x if mirrored. Typically, you purchase enough memory so that swapping never occurs and only configure 512MB-1GB of primary swap. If you run in this mode, you must enable pseudoswap by setting swapmem_on=1. My ERP systems have not swapped at all in over four years. Having said this, I would still opt for generous internal disks. The idea is that you have two mirrored boot disks on separate controllers so that there is never a need to shutdown for disk replacement. If you choose the minimal swap model (and I suggest that), you should configure separate dump areas that are not mirrored. I just think it is rather dumb to buy a ton of disk that you are never going to use. Bear in mind, if you adopt the small primary swap model you can always add additional swap at any time if needed.

[Paula J Frazer-Campbell]

Hi

I would go for two 36 Gig disks.

Build the server on one and then mirror the 2nd in.



As well as swap allow room in /stand for additional copy of vmunix as well as vmunix.prev.

This might be helpful:-

To mirror an existing root volume:
1. pvcreate -B /dev/rdsk/device (-B creates BDRA area)
2. vgextend /dev/vg00 /dev/dsk/device
3. mkboot /dev/rdsk/device (put boot utilities into boot area)
4. mkboot -a "hpux boot arguements here" /dev/rdsk/device
5. lvextend -m 1 /dev/vg00/lvol1 /dev/dsk/device
6. lvextend -m 1 /dev/vg00/swaplv /dev/dsk//device
7. repeat lvextend in sequence for all root logical volumes
8. lvlnboot -v ( to update the boot information with the new disk)


Paula

[James R. Ferguson]

Hi:

An 18GB or 36GB drive is fine for vg00. Only the standard operating system filesystems belong on vg00, in my opinion. I'd only place a primary swap and dump space on vg00. Additional (secondary) swap space should reside on another physical disk for best performance.

Regards!

...JRF...

[S.K. Chan]

As far as the server goes you can have up to 4 internal hot plug disks. You got 2 choices ..
a) 36GB hotplug ultra scsi (A6831A)
b) 73GB hotplug ultra scsi (A6725A)
The part numbers are quoted from what I have, they might be old already. Like what others have said you no longer need so much swap to match your memory. Keep primary swap to a minimum 1GB and added additional swap when needed. It's best to separate the additional swap on a different disk than the root disk. Having said all that, what's the harm of allocating 12GB of primary swap (at 1:1 ratio) space since the disk capacity nowadays is huge apart from wasting all the space .. that way you don't have to worry about separate dump space, etc, etc. Just my thought ..

[Todd McDaniel_1]

I work for a large telecom company and our policy based on recent data, is to allocate around 30% to 50% of GBs of memory for swap space. Depending on load of course, but in general the lean towards 35 to 40%.

A few of my other boxes have only 25% swap vs memory.

One of my boxes has 54GB of memory and only 26GB of swap.

I have been told that this is the new standard by my System intro team and HPCE onsite.