Controlling primary/alternate paths under 11.31?

[Kevin (Gonzo) Bushman]

We have a NetApp for storage. We have to 11.31 systems and two SAN switches for redundancy. One switch connects to each system as the primary path and then to each of two filers in the NA (with two paths per switch for each system). The second switch for each system works as an alternate path in the event the primary switch for one system was to go down.

The problem I'm finding is that under 11.31, it is balancing the I/O load across all 4 paths (2 primary and 2 alternate). Due to the layout of the cables, we don't want to use the alternate path unless a switch is out for some reason. Can this be controlled? And if so, how?

BTW - all my VG's are set up with agile naming.

[Torsten. · Accepted solution]

Why use the second switch as standby only?


However, the server will do round_robin per default:

# scsimgr get_attr -N /escsi/esdisk -a path_fail_secs -a load_bal_policy


You can change by using scsimgr (see manual or man page).

[Steven E. Protter]

Shalom,

scsimgr lets you control the policy.

Policy impact:
round robin is easy on the system. The kernel does not need to do work to make that happen.

http://bizsupport1.austin.hp.com/bizsupport/TechSupport/Document.jsp?lang=en&cc=us&taskId=120&prodSeriesId=417289&prodTypeId=18964&objectID=c01685302

Least load requires more kernel work but has been known to provide substantial performance impact.

SEP

[Kevin (Gonzo) Bushman]

We use the second switch as an alternate because we actually have two filers in the NA. The second switch is actually connected to the other filer. Which is what is causing the problem. Data going down this path goes to the second filer, goes across the back plane of the NA to the other filer, that filer serves up the data which then follows the same (longer) path back to the O/S.

What I want/need to do is turn this off or set it purely as an alternate so ALL I/O goes only down the primary path. Unless of course there's an issue with the switch, when if that is the case, it needs to down the alternate path.

Make sense?

Thanks. I'll look at scsimgr.

-G

[Kevin (Gonzo) Bushman]

Steven -

Thanks for the link. That is most helpful!

Points given.

I'll look at the links and suggestions. I knew there was a solution somehow!

-G

[Torsten.]

Does your array support ALUA?

In this case the server should be aware of this and adjust automatically.

[Torsten.]

Get the complete configuration:

# scsimgr get_info -D /dev/rdisk/disk__