NewBe config: SG A.10.20 -- Cannot properly create VG with CDFS on both nodes

[chisle]

Hi!

I have a vg: /dev/vgSGapps

I have created the cluster already, and clquery shows up, etc., and has package sgeasy (of course) as I used easy deploy, so of course I've already done the dfs config on both nodes, and we're not using legacy anywere (that has already been turned off).

Now, /dev/vgSGapps has only disk that are shared dsf (/dev/cdisk/disk*), and I built it that way on purpose, as it is easier to manage the disk on both nodes. It was built out before configuring SG on the box.

Following the manual instructions, I have done the following:

1. vgchange -a n vgSGapps
2. vgexport -p -s -m /tmp/vgSGapps.map vgSGapps
3. copied the file to node2
4. on node2, after creating /dev/vgSGapps and the mknod command (c 128 0x003000, to match node1), I used vgimport -s -m /tmp/vgSGapps /dev/vgSGapps
5. I verified that it pulled in disk, but Lo and Behold: it is not using CDFS, it is using DFS!!!

So, to rectify, I tried adding the CDFS devices then remove the DFS devices, but that didn't work...it complained that the disk were in use (of course, it knows that the file systems that were on node1 are on those disk...but for some reason it isn't mapping propelry back to cdfs, eventhough ioscan -fnkNC disk /dev/disk/diskX points to the proper /dev/cdisk/diskY device, shared between the two nodes)

So, if you are using CDFS, what is the proper way to do this?

Thanks!

[Michael Steele_2]

I'll start researching for you but I'd like to know if you have the same ignite image on both nodes are noted in MC/SG, the same patches, etc.

You can get a good idea :

swlist -l fileset | wc -l
swlist -l package | wc -l
swlist -l patch | wc -l

...on all nodes int the cluster.

And if they don't tally, then here is where I would start. I'd take an ignite image of the primary and reignite the alternates with it. It won't take long and puts you were MC/SG wants you to be.

[chisle]

yes, the images were created from a Gold that I created. Also, the same patch clusters were put down, etc.

Furthermore, swlist is identical, swverify comes back clean, etc. and the chck for the cluster setup as well as SGInfo all come back squeaky clean.

[chisle]

Found it!

Needed -C option....

so, instead of:

vgimport -N -s ....

use:

vgimport -N -C -s....

Thanks!

[chisle]

-C to the vgimport command fixed the issue.