Re: MPIO disks and paths

On Mar 6, 7:52 pm, David J Dachtera <djesys...@xxxxxxxxxxxxxxxx>
wrote:
b...@xxxxxxxxxxxxx wrote:

Cam is right about that kit from HP. You must load that kit from HP
which provides ODM entries to define the default attributes and
changeable attributes for the EVA disks. It also provides a few HP-
provided commands like lshsv, lshba, rmhsv and hsvpaths.

By default, when AIX scans the adapters, etc during a cfgmgr it
enumerates the disks in ascending LUN order through the first adapter
it finds which explains why all the disks are on the same (lowest fcsx
number) physical path upon a reboot. Likely some cfgmgr details
missing here.....

If you download the 1.0.1.0 HP MPIO kit, then you can use path
priority to load balance. The AIX 5.3 System Management Concepts:
Operating System and Devices book has a little info on path priority
in the MPIO section. The highest priority is 1 so any path set to a
priority of 1 would be the preferred path. You can effectively load
balance using this feature. Also you could set the path used next
for failover.

Another method that works sometimes is to use the chpath command to
disable the current path. This will force the hdisk to use an
alternate path, typically the next path instance in the list. Once
the disk has failed over, you can enable the previously disabled path
and you now are using a different path and likely a different adapter
since the path instances typically alternate between the adapters.
However, *sometimes* when you enable the previously disabled path, the
hdisk will go back to that path so you have accomplished nothing.

The last method I know of is to set a perferred path for the vdisk in
Command View EVA on the SAN Appliance. If you choose a preferred
path, choose only the paths that are listed as "Path A Failover Only"
or "Path B Failover Only". This would at least allow you to have the
hdisks spread evenly across the available adapters but not across all
the available EVA ports.

Pardon my butting in here...

My shop will soon be moving from VMS to AIX. So, I'd like to pose a couple of
quick questions.

From what you've written above, I take it that upon a "mount" command AIX does
not attempt to switch the (disk) being mounted to the path with the least
currently "mount"-ed (volumes). (Sorry - some VMS terminology does not translate
directly to UN*X.)

Is this a correct conclusion?

By contrast, when we MOUNT our FC disks in VMS, the system will auto-switch them
to the least-used path of the available paths to each device at the time we
issue the MOUNT command.

Then, regarding manual path switching, does it take a third-party supplied
command utility such as the mentioned "chpath", or is something supplied with
AIX to achieve this?

The VMS equivalent would be SET DEVICE/SWITCH/PATH=PGcu:wwid device_name
(In VMS's command language, command keywords/parameters are separated by white
space, qualifiers are delimited by slash ("/").)

Also, do you perchance have any experience with EMC and AIX? Does EMC's
"PowerPath" provide utilities similar to what you mention from HP?

Forgive my AIX newbie-ness. I did UNIX back in 1986 for a year or three, and
more recently FreeBSD and Linux on PCs, but I've been doing VMS since 1983.

--
David J Dachtera
dba DJE Systemshttp://www.djesys.com/

Unofficial OpenVMS Marketing Home Pagehttp://www.djesys.com/vms/market/

Unofficial Affordable OpenVMS Home Page:http://www.djesys.com/vms/soho/

Unofficial OpenVMS-IA32 Home Page:http://www.djesys.com/vms/ia32/

Unofficial OpenVMS Hobbyist Support Page:http://www.djesys.com/vms/support/- Hide quoted text -

- Show quoted text -

The only thing I'll add is that using HP's XP1024 with AIX appears to
be different then using HP's EVA storage. We have an XP1024. Sorry I
haven't used EMC storage. It was SSA (which I really liked) and a
FAStT before the XP1024.

The XP1024 driver or PCM does NOT support load balancing or robin
round. All you get is path fail over. Hence the efforts I went through
when I realized all my disks were on the same FC adapter, to try and
balance the load. Now I can read / write about 130-150 Mbytes/s
aggregate across two FC adapters. Much better performance.

.