Force OS to recreate I/O device trees

[Alzhy]

Our 11.11 environments are so large -- over a thousand device paths (EVA, XP, etc) and from time to time -- we experience those WSIO bus exceeded 255 errors. The workaroudn always have been a teious process from HP.

As I am getting rusty (and can even hardly recall the trics of the trade) -- what is th best way to accomplish this?

I am thinking:

remove /stand/ioconfig and /etc/ioconfig?
Reboot?

Our Storage are all VxVM based so it is agnostic to cXtYdZ changes. OS is however LVM


TIA for any insights.

[Duncan Edmonstone · Accepted solution]

Basically yes, but preparation is key to continuued survival!

Remember this won't just change disk devices, but LAN devices to (and if you have any esoteric devices such as MUX's or PSI cards attached, well I guess they might change to)

So make sure you have a full "ioscan -fn" and "netstat -in" saved before you begin, mnake sure you have an Ignite backup and an OS recovery disk, then go ahead and remove and reboot to LVM maintenance mode where you can run "ioinit -c" to recreate the files if required (can't remember if /sbin/ioinitrc gets run in LVM maintenance mode - probably does). Then you'll need to export/import your root VG and recreate root/boot/swap/boot bindings with lvlnboot, and then check if your LAN instances have changed and adjust netconf appropriately.

Incidentally, I can understand having a lot of LUNs presented from an XP, but from an EVA? surely you only need a few?

HTH

Duncan

[Alzhy]

Duncan, yes sir. Our ecosystems are fairly large and are still 11.11. Our EVA systems are used as BusinessCopy targets for our XP storage (External storage XP) so we provision one for one. But really my issue is on our rather large non-prod servers where we present a very large number of XP and EVA disks that represent all our production ecosystems. And it is very dynamic, we often move environments around.

And how we do it is rather simple but I think needs a tweak due to these problems. Since our boot disks are SAN Boot disks, we simply "resettle the OS disk" on the target server/partition where they may already be existing LUNS presented but at a different luin number range or ranges. Normally when these WSIO errors occur, we employ HP's cure but it is getting to be a curse. So I am thinkning maybe we would want to incorporate "cleaning" up the I/O tree after each environment move.

We are very well versed with tracking where our hot network drops are as well as our APA pairings so that is not an issue.

Thanks though.