HP-UX 11.11 Behaviour with Storage Array Move from One SAN Switch to Another

[Alzhy]

We've migrated a big XP and several EVA from a Big SW12K to a SW48K SAN Switch infrastructure. Our environments are mostly HP-UX 11.11 (and several Solaris and Windows/ESX). The migration was done COLD - all servers were down 'cept for 2 SandbBox HP-UX 11.11 machines.

Lo and behold, the WIndws and SOlaris ecosystems did not have problems on startup! BUT our HP-UX ecosystems had many problems to wit:

1.) On our very large systems with hundreds to thousands of disks and disk paths, we encountered WSIO (bus exceeded) problems. It seemed HP-UX on startup needed to establish new device/bus instance files on the system. Luckily we've a method to cleanse our way out of this hole but it requires a lengthy reboot

2.) SAN Boot Paths changed coupled with the XL2 HBAs issue with the SW48K

3.) Our LVM managed servers had problems due to changes in CxTyDZ values... luckliy we save VG Mpfiles on a regular basis. NO problems of course with the majority of our sevrers that are managed with VxVM as VxVM cares not about CxTyDZ values.

Anyone care to comment on what we should have prepared on the HP-UX side of things? Is there a way to figure out the new disk paths when ever Disk Storage Arrays migrate to a new port or switch? Anyone expeirenced a similar issue with WSIO bus exceeded errors?

What's unbelievable is tht Solaris (8 at that) and WIndows survived our exercise. I know that the issue most liely is fixed in 11.31 but should there have been a way for us to prepare better?

Thanks.

[Duncan Edmonstone]

Nelson,

No, not really - the IO stack was really a weak link in v1/v2 and as you already pointed out it got fixed in v3.

When FC disks started appearing there was really 3 ways to go on how disks were addressed:

i) By FCID (the way HP-UX and AIX went)
ii) By WWN (the way Tru64 went)
iii) By WWN with some sort of lookup and method of persistence (the way Solaris/Windows went)

In this case I guess MS/Sun got it right and HP didn't - all we can say is "its fixed now" and its yet another good reaon to upgrade to v3

I guess if you had to replace a storage port due to a failure and the new device got a new WWN, then Solaris/Windows would have problems whilst 11iv1 wouldn't so to some degree its swings & roundabouts...

HTH

Duncan

[Duncan Edmonstone]

Nelson,

WRT point 1 above, I presume you were hitting the maximums described in this WP here:

http://docs.hp.com/en/7154/Max_LUN_whitepaper_v11.31.pdf

And had to regenerate the ioconfig files from scratch (quite painful from memory...)

Did you see the note about driver and patches increasing the maximum LUNs in 11iv1 from 4096 to 8192 - would that have helped or were you already at the 8192 max?

HTH

Duncan