help cluster problems

[mohammed arafat_2]

Dears
i have this problem

i need to distribute my VG LOCK on onother node i've tried by using lvm command line but when booting the servers this disks on DA2300 go to scan mode thery are not claimd on both servers after i do ioscan -fnC disk

i have 2 rp8400 and 1 DA2300
i want to make cluster with mcsg
i've made all the steps on the mcsg manual chapter 5
and most my problem happens when i export import my vg that will be the cluster lock

so after all the problems i had
i've decide the following
1) i've configure vg01 on node1
2)i've configured the cluster containg only node 1
3)on SAM ---volume groups---Action--distibute
was active

so when i wanted to ditribute vg01 to node 2 it gave me an error that network failure or rpc failuer
i've checked the network it is fine

how can i resolve my rpc problem to make vg sharable on both node to continure cluster configuration
thank you

[Karthik S S]

Hi,

I am not sure what procedures you are following to distribute the cluster lock VG to the other node.

Well to give you an rough Idea,

On Node1:
1. pvcreate /dev/dsk/cxtxdx
2. mkdir /dev/vg01
3. mknod group c 64 0x010000 (Make sure that on the other node also you give the same minor number when u export the VG01)
4. vgcreate /dev/dsk/cxtxdx vg01
5. (Create required lvols on this VG)
6. vgexport -p -v -m /tmp/vg01.map vg01
7. ftp this map file (/tmp/vg01.map) to the 2nd node)
8. vgchange -a n vg01

On Node2:
1. mkdir /dev/vg01
2. mknod group c 64 0x010000
3. vgimport -v -m /tmp/vg01.map vg01 /dev/dsk/cxtxdx
4. vgchange -a y vg01
5. vgdisplay vg01 (to verify)
6. vgchange -a n vg01

On Node1:
9. vgchange -a y vg01
10. cmquerycl -v -C /etc/cmcluster/cluster.asc
11. Now you should see vg01 in the cluster.asc file.

Regards,
Karthik S S

[Karthik S S]

Also when you power on the systems,

1. make sure that the DA2300 is properly cabled to both the RP8400 systems.
2. Power on RP8400s
3. Check the o/p of "dmesg" to see if there are any SCSI errors.
4. Run insf -e -C disk
5. And on both the servers run ioscan -fnC disk to see if the disks are claimed properly.

-Karthik S S

[mohammed arafat_2]

what is the difference between
vgimport -v -m /tmp/vg01.map vg01 /dev/dsk/cxtxdx
and vgimport -s -m ,,,,,

s stands for share and its for making the vg sharable between the 2 noods

i have a problem ib sharing

[Karthik S S]

Oops ... sorry about the multiple postings. But forgot to add that when you connect a Disk array to two systems you also have to make sure that SCSI IDs for the SCSI controller on both the systems are unique. If there is a conflict in the SCSI bus disks may not get claimed.

- Karthik S S

[Karthik S S]

the -s switch option in vgexport/vgimport command enables the VG to be activated in both the systems simultaneously. But, when you configure a lock VG this is not the recommended configuration. Because, either of the system has to exclusively own the lock VG.

-Karthik S S

[mohammed arafat_2]

can you help on that Karthik

how can i check if i have a conflict on the scsi ids , i ahve 2 scsi lvd i dont know the p/n but there is no dip switch on them

plz help

[Karthik S S]

I mean you should check the SCSI IDs of the SCSI controller cards on the RP8400 systems. From the BCH prompt you can do that.

From the main menu of BCH (Main Menu: Enter command or menu >), enter

SER
SCSI

The output will be something like this,

Path (dec) Initiator ID SCSI Rate Auto Term
------------ -------------- ---------- ---------------
0/0/1/1 7 No Limit Unknown


In the above example the SCSI ID of the controller is set to 7 which is default.

So, in one of the RP8400 you change this value to 6.

HELP SCSI

will show you how to do this. (From Service menu)

-Karthik S S

[mohammed arafat_2]

thank you Karthik

i've submited a 9
if it worked i will give 10000000000

thank you man

[Karthik S S · Accepted solution]

Hi,

Pl. let me know if it worked ... I am not asking this to get 100000000 points .. but I am curious to know if my answers have helped you in resolving the problem ;-)

-Karthik S S

[mohammed arafat_2]

it worked man , thank you very much