Disk device files are not the same on 2 node cluster

[Brian Ruisi]

I hope someone can give me a hand. I am setting up a 2 node cluster. I initially had a bad Fiber cable configuration on initial OS install. The two nodes share a JBOD and a Disk Array. I have the cabling corrected so that both nodes should see each disk device through the same path (0/8/0/0..., etc...). The problem seems to be that each server assigned different device files for the same disks (C12T0D0 on node 1 and C6T0D0 on node 2, etc...). After fixing the cabling issue, I deleted all related files in the /ETC/DSK and ETC/RDSK directories and ran INSF -e. The same assignment came up again. Is there something I can do to make the system reassign new device files, and then they should show the same on both nodes? Thanks in advance.

[Armin Feller]

Hi,

perhaps a 'rmsf -H ' and 'insf -H ' will fix this simply.

If not, please have a look here:

http://forums.itrc.hp.com/cm/QuestionAnswer/1,,0x46cb50dde50cd71190050090279cd0f9,00.html

Regards ...
Armin

[melvyn burnard]

there is no REQUIREMNENT for the device files to be the same.
Are the systems physically identical, and paths the same?
The device files are created based on the card instance numbers as known by the system.
if you look at iosan and check the instance number for the card the drives are connected to, htis will give you the C? number.

If this is not what you need, then you are into playing with the ioconfig file, which I do NOT recommend.

[John Poff]

Hi Brian,

The other side of the argument is that you really don't need the device files to be the same on both nodes. It is nice to have it set up that way, but if you understand your layout and have things documented well, it really doesn't matter so much.

We did the same thing here when we setup our first cluster. We went to great lengths to get our two V boxes to have all the cards in the same slots and addressing all the devices the same way. Then we added a third node with a completely different configuration and it shot the identical part to pieces. Now, after having put together several clusters, it seems like more trouble than it is worth to scramble around and get everything identical.

Just my two cents worth.

JP

[Uday_S_Ankolekar]

That's OK if you have different device name on same path , In case you want to have same device name on both node then you have to play little bit with ioscan , ioinit .!
There is a good doc available (KBRC0004879) to change the device name , Basically these are the steps:
ioscan -f | grep -e INTERFACE -e DEVICE | grep -v target | awk '{print $3, $1, $2 }' > infile . This creates a file infile.
Change ext_bus and lan instance as per other server, Run same on next node and rcp that file on first node. Run /sbin/init/init 1 ; sbin/ioinit -f /infile -r This will reconfigure kernel and reboot the box. ran insf -e
Now ioscan will show disk instances same on both server