HPE 9000 and HPE e3000 Servers
1827277 Members
1947 Online
109717 Solutions
New Discussion

Question in understanding Hardwarepathnames

 
SOLVED
Go to solution
Lothar Krueler
Regular Advisor

Question in understanding Hardwarepathnames

Hi there,
Two L-2000 servers are connected each to the X and Y Controller of a Autoraid 12H. On each server the hardwarepaths 0/4/0/0.1.[0-7] and 0/7/0/0.0.[0-7] are leading to the LUNs.
If I look at the servers backside it seems that both machines are connected in the same manner to the X / Y Controller. Is it right , that 0/4/0/0... from each server are connected to the same controller (say X) and the other hardwarepath is connected to the other (say Y) controller ?
I must understand that for optimizing performance.
Thanks in advance,
Lothar
Wissen macht zaghaft, Dummheit kann alles!
5 REPLIES 5
A. Clay Stephenson
Acclaimed Contributor
Solution

Re: Question in understanding Hardwarepathnames

Hi Lothar,

In this case how they are connected physically is not a problem. It is how the volume groups are configured (which is the primary path to each LUN and which is the alternate path to each LUN that matters). The idea is to mix the primary path on the LUNS.

Here is an example,

Let's say that on both machine A and machine B have 2 devices that both point to LUN0 :/dev/dsk/c2t0d0 /dev/dsk/c3t1d0
t0 is AutoRAID controller X; t1 is Y and note they are on separate SCSI busses c2 & c3
but both actually connect to LUN0 for redundancy.

We also have on both machines /dev/dsk/c2t0d1
and /dev/dsk/c3t1d1 that point to LUN1.



After you have done a pvcreate on each LUN,
you then vgcreate (or vgextend for additional drives/LUN's). The first device path your list is the primary and the second is the alternate.

On machine A:
vgcreate /dev/vg03 /dev/dsk/c2t0d0 /dev/dsk/c3t1d0

on Machine B:
vgcreate /dev/vg03 /dev/dsk/c3t1d0 /dev/dsk/c2t0d1

The effect of this is that Machine A's primary path is through X and Machine B's primary path is through Y.

Now for best performance you really should extend this concept. If you need a device of 100GB; create 2 LUN's of 50GB each. Use both of these in a volume group with alternating primary paths. Each logical volume should be stripped across both LUN's.

You can do this for each volume group and then both machines use both AutoRAID controllers all the time. It took me a while to develop this technique (and doing it SAM or for ServiceGuard is tricky) but it does really make an AutoRAID perform. It's actually easier to configure like this from the command line like real UNIX guys.

The other point is leave as much space as you can unconfigured.

Regards, Clay
If it ain't broke, I can fix that.
Bill McNAMARA_1
Honored Contributor

Re: Question in understanding Hardwarepathnames

> Two L-2000 servers are connected each to the
> X and Y Controller of a Autoraid 12H.

bus 1
0/4/0/0 (SCSI ID 7)
L1 ----------- X
/
/
L2 ---------
0/4/0/0 (SCSI ID 6)

bus 2
0/7/0/0 (SCSI ID 6)
L1 --------
L2 ------------ Y
0/7/0/0 (SCSI ID 7)

Use primary for 1/2 luns throug L2 0/7/0/0
and alternate on L1 0/7/0/0

For the remaining luns access throu L1 on 0/4/0/0 with alternate on L2 0/4/0/0

Doing this will help keep switchover device files similar.
(ServiceGuard VG activation and package configuration dependant)


The HW paths
0/4/0/0 ext_bus (scsi card)
.1 SCSI ID of Autoraid controller Y
.[0-7] LUN

0/7/0/0 ext_bus (scsi card)
.0 SCSI ID of AR controller X
.[0-7] LUNs.


>Is it right , that 0/4/0/0...
>from each server are connected to the same
>controller (say X) and the other hardwarepath >is connected to the other (say Y)
>controller ?
Make sure that the 2 controllers have different high priority scsi ids 7 and 6
Try accessing through the 7 as primary.
Remember there will only be one bus with the X controller and the two initiators (scsi cards).. thus they must have different scsi ids.


Later,
Bill
It works for me (tm)
Lothar Krueler
Regular Advisor

Re: Question in understanding Hardwarepathnames

Hi Clay, Hi Bill,
thanks for your fast responses.

Bill, I'm a little confused about your resply, think I don't unstand it.

Clay, one of my problems was the restriction of max 8 LUN each 12H, but now i found a way to save LUNs, so that I can use 2 LUNs each VG.

zeus: 0/4/0/0.1.0 c6t1d0
0/7/0/0.0.0 c9t0d0

hera: 0/4/0/0.1.0 c5t1d0
0/7/0/0.0.0 c7t0d0

Is it right, that the "t1" at zeus indicates physicaly the same controller as the "t1" at hera ? In this case the answer to my question is done.

I think it's a good idea to follow your concept - that's the reason for my question.
Unfortunately I have a little difficulties in English, so configuring a striped pair of LUNs as a MC/SG Volumegroup is not really clear to me. I should not use SAM, using the command line would be easier, right ? Is there a cookbook for doing this ?

Thank you both,
Lothar
Wissen macht zaghaft, Dummheit kann alles!
A. Clay Stephenson
Acclaimed Contributor

Re: Question in understanding Hardwarepathnames

Hi again Lothar,

I'm certain that your English is much better than my German. You were lucky, in that I had a text document which explains this setup. I first created it for my other admins. It talks about a ServiceGuard cluster but other than doing a vgchange -c y the instructions are the same.

Hope this makes it clearer, Clay
If it ain't broke, I can fix that.
Lothar Krueler
Regular Advisor

Re: Question in understanding Hardwarepathnames

Hi Clay,
thank you very much, this is much more I expected!
This weekend I'll start holidays for 3 weeks, after that I`ll try the configuration.
:-))) Lothar
Wissen macht zaghaft, Dummheit kann alles!