From: Chris Gregors (Chris.Gregors@TELUS.COM)
Date: Tue May 13 2003 - 15:28:12 EDT
OK. I'm wrong on the first point. The pvid's should match between the
hdiskpower[x] devices and the hdisk[x] devices that make up the
hdiskpower[x].
I was thinking of the EMC serial numbers that show up in an inq. They
differ because the paths to the disks are different. This would be the
same on scsi or fiber.
Unfortunately you cannot see the pvid's for the hdiskpower and hdisk
devices at the same time. Since only one or the other can be in a vg at
any given time.
I just gave our EMC CE a call and he informed me that the requirement of
installing powerpath on all nodes on the same FA/SA has been removed in
recent firmware code changes on their end.
The only concern I have with mixing powerpath and non-powerpath in a
cluster is the potential for a mistake when allocating disks.
-----Original Message-----
From: Bruce Zimmer [mailto:b.r.zimmer@WORLDNET.ATT.NET]
Sent: Tuesday, May 13, 2003 12:30 PM
To: aix-l@Princeton.EDU
Subject: Re: HACMP and EMC powerpath
I will disagree on two points, and that is that powerpath is not
required on all nodes, and that the PVIDs on the hdisk & hdiskpower
devices are the same. We ran quite successfully on a pair of S7A's
(cascading) where our primary node used power path and the secondary did
not. The PVIDs were the same on hdisk & hdiskpower devices. The
hdiskpower devices are simply parallel paths to the same physical disk
in the EMC Symmetrix box. This installation was SCSI only so if fibre is
involved, I may be all wet.
Bruce Zimmer
Central Data Systems
(248) 615-4644 (direct)
(248) 320-1175 (cell)
bzimmer@centraldata.com
-----Original Message-----
From: IBM AIX Discussion List [mailto:aix-l@Princeton.EDU] On Behalf Of
Chris Gregors
Sent: Tuesday, May 13, 2003 12:11 PM
To: aix-l@Princeton.EDU
Subject: Re: HACMP and EMC powerpath
>From my experience and memory of discussions with EMC CE's
1. Powerpath creates (semi)virtual devices that encompass the 2(or more)
real disks behind the scenes. These new disks are called
hdiskpower[x]
2. The hdiskpower[x] devices have a different pvid from the hdisk[x]
devices.
3. You should make your vg's using the hdiskpower[x] devices only! If
you
mix them you can end up in a strange place.
4. Both nodes of the cluster will need powerpath. If you put it on the
primary
node only, then you cannot use the hdiskpower[x] devices in the
failover mode.
Therefore you cannot use the hdiskpower[x] devices on the primary
node,
therefore you don't need to unstall powerpath.
5. If this is fiber disk (vs scsi), you MUST install powerpath on all
nodes that
live on the same FA from the EMC frame. This is a restriction of how
EMC does
stuff.
Otherwise it works quite nicely. The key command you will get to learn
is:
/usr/sbin/powermt display dev=all
Later.
Chris.
-----Original Message-----
From: Myers, Drew [mailto:MyersD@GOALAMO.COM]
Sent: Monday, May 12, 2003 8:58 AM
To: aix-l@Princeton.EDU
Subject: HACMP and EMC powerpath
Hi all,
I have a few questions regarding HACMP and (shocking) powerpath.
Currently, we have a 2-node HACMP cluster arranged in a "rotating"
cluster configuration. This has been tested and each node is able to
takeover the resource group just fine, with no issues.
Now I've been asked to install powerpath onto one node only (the
production node). I'm still reading through the powerpath docs, as I've
never done this installation before.
If you have any pointers, gotchas, etc., I'd like to hear about them.
Thanks,
Drew Myers
perotsystems
(954)769-2489
This archive was generated by hypermail 2.1.7 : Wed Apr 09 2008 - 22:16:50 EDT