Native Multi-Pathing in HP-UX 11.31

[Jeff_Traigle]

Seems getting an answer via support channels is taking an inordinate amount of time so I thought I'd ask here for a hopefully quicker insight.

HP published a white paper titled "HP-UX 11i v3 Native Multi-Pathing for Mass Storage," (http://docs.hp.com/en/native-multi-pathing/native_multipathing_wp.pdf) which explains some of the workings. It misses a few key details, however.

Our application is very sensitive to timeouts (10-30 seconds) so we need to reduce the time it takes for an I/O to timeout and be sent via a different path. On page 12, the white paper indicates that either a time based or count based policy is used for I/O retries. Their is no mention which is the default policy and how to change to the other policy if desired.

After some digging, I found the retry_delay_enabled setting in the scsimgr man page. It's set to true by default. Is this setting what controls which I/O retry policy is used? If so, what are the pros and cons of using the count based policy instead of the time based policy? (Count based would seem to eliminate our timeout problems if a path goes down, but not sure if there's some gotcha I'm not thinking of in switching to it.)

Anyone know the nuts and bolts of this feature who can shed some light on a foggy brain?

[Olivier Masse]

If you don't get anything concise from the response center and have a good relationship with your HP rep, you can PM me at omasse at mayoxide dot com. I swapped a few business cards last year with some people who were involved in the stack's design. Although I can't give you these contacts directly (and they might not be at liberty to answer public questions anyway), I can type up the exact name of the lab at HP so you can leave it up to your rep to officially locate some top people that might be able to provide answers.

I'd be interested in whatever you find, by the way. I'll monitor this post, please share any findings you get.

Olivier

[Jeff_Traigle]

It took a while, but I finally got the answer from HP in October. I wanted to take a minute to post the answer in case anyone else comes across this post looking for similar information.

The infinite_retries_enable parameter controls which I/O retry policy is in effect. Time-based is the default (infinite_retries_enable=false). To turn on count-based I/O retry policy, this parameter must be set to true.

To see what the current setting is:

scsimgr get_attr -N /escsi/esdisk -a infinite_retries_enable

To make the I/O retry policy count-based (and persistent between reboots):

scsimgr save_attr -N /escsi/esdisk -a infinite_retries_enable=true

Thanks to Steve Lincz who finally was assigned the case and got me the answer I needed.