11.31 changing IO timeouts for an online EVA upgrade

[Olivier Masse]

I'm planning to upgrade online two EVA8000s which are hooked up to a mix of 11.23 and 11.31 HP-UX servers.

In the past, I had to follow closely the paper "HP StorageWorks Enterprise Virtual Array
online firmware upgrade best practices" which instructed how to set disk IO timeouts to 60 seconds prior to the update using "pvchange -t 60 /dev/dsk/xxxxx". I did this, the EVAs took maybe 40 seconds to restart, and I didn't have any problems.

However, that guide doesn't seem to have been updated explicitly for 11.31. It's not even available from the BSC when looking at the EVA manuals - I had to make a Google search to locate it.

My concern is that with the new storage stack, I'm not sure if doing a "pvchange -t 60" will be enough. For example, if I do this on a PV, and then run "scsimgr get_info" on the disk directly, it sill mentions that the IO transfer timeout is at 30 seconds.

I was wondering if some here went through the process and would have some suggestions.

For those who might ask, XCS 6.220 is available from the software depot and that release seems to be customer-installable, so I am evaluating the possibility of doing it myself to save the hassle of having to schedule someone from HP to come in and do it.

Thanks

[Stephan.]

Hi,
try http://docs.hp.com/en/scsimgr/scsimgr_whp_AR0709.pdf starting at page 16 to get an idea how scsimgr handles tunables.

Personally after my last experiences with eva firmwares i wouldn't suggest to do it alone even more if there is no explicit need and your current version is running smoothly.

[Olivier Masse]

The fact is, it's not that smooth. We have outstanding bugs on two of them that require controller resets, and while I'm at it I'd like to perform an upgrade.

Thanks