Adding A Disk To A Volume Group In A ServiceGuard Package

[John Jayaseelan]

Hi,

I need to add a disk to the Volume group under Service Guard cluster. Should bring the VG out of cluster to add the disk or Disk can be added straight forward. The volume group disks are arranged in HASS array where I have free slots at the bottom of the array.

Please tell all the steps required to implement it.

Thanks
John Jayaseelan

[Vincent Fleming]

John,

This is really a MC/SG question, so I'm moving it to the HP-UX ServiceGuard forum - it's much more likely to get answered there...

Vince

[Sridhar Bhaskarla]

Hi John,

You do not need to bring down the cluster or package to add another disk. Add it as if you are adding to a stationary volume group.

However, you will need to export (preview) and import it on the secondary node to get the changes distributed.

On the primary node
#pvcreate (-f) /dev/dsk/c?t?d?
#vgextend vg?? /dev/dsk/c?t?d?
#lvcreate -n some_lvol -L size /dev/vg??
#newfs -F vxfs /dev/vg??/rsome_lvol
Edit /etc/cmcluster/your_package/package.cntl
Add the new logical volume and file system into it.
Copy package.cntl onto the second node

Note down the group minor number of vg??

#vgexport -p -v -s -m /tmp/vg??.map vg??
Copy /tmp/vg??.map to the secondary node into /tmp

On the secondary node

#vgexport vg??
#mkdir /dev/vg??
#mknod /dev/vg??/group c 64 0x0?0000
#vgimport -v -s -m /tmp/vg??.map vg??

Replace "?" with your numbers|names.

You are done.

-Sri

[Byron Myers]

One addition to Sridhar's directions - before running pvcreate. After inserting the disk, run ioscan, then "insf -e" so the system will make the device file for the new disk. Do this on all SG systems attached to the disk.

[Brian M Rawlings]

Sri (& John):

I am under the impression that you must do a cmapplyconf , to create and distribute the cluster binary file to all nodes, any time you make a change to a cluster config file.

Is this step not necessary? You ARE making a significant change to a package config file, and one of the args to cmapplyconf is a -p for each package.

Other than this question, the above is spot on, good explanation. Hope you can respond to this question to get it put to rest, one way or the other.

Regards, --bmr

[John Poff]

Brian,

The only changes that would be made are in the package control file. You don't need the cmcheckconf/cmapplyconf part unless you make changes to the package config file.

JP

[Sridhar Bhaskarla]

Hi Brian,

We are adding only a disk to a volume group. This does not require changes to either the package configuration file or the cluster configuration file.

Adding an extra logical volume does need an entry in the package control file but that doesn't require cmapplyconf either.

The only case related to disks where we would recompile the cluster configuration is when you
change the cluster lock disk.

-Sri

[John Jayaseelan]

Hi All,

I am adding the disks to enable mirroring on existing LV, so I would be creating any new Logical Volume or file system.

I think, I do not need to do any change on the package. Could you please confirm it?

In addition, I am adding the disk into an available slot in the HASS Unit. Apart from setting the unique jumper settings what else need to be done to successfully import this disk.


Thanks
John Jayaseelan

[Ashwani Kashyap]

If you are just adding the disk to mirror an LV in the cluster the you don't have to edit any package file .

Just add you disk with a unique ID . Do an ioscan followed by insf -e .
The do your pvcreate , vgextend etc and then mirror the LV .

However you still will have to export the VG and import it to all other nodes as sridhar explained earlier .

[John Jayaseelan]

Thanks for the help. I will try and update if there any issues.

John Jayaseelan

[John Jayaseelan]

Hi,

Is there any update on this issue please.

Thanks

[Robert-Jan Goossens]

Hi john,

take a look at next question.

Robert-Jan.

[Robert-Jan Goossens]

Oops

http://bizforums.itrc.hp.com/cm/QuestionAnswer/0,,0x880257bd90a9d611abdb0090277a778c,00.html

[John Jayaseelan]

Hi,
I want an update for the following.

After adding the disk I was getting a message very often

'Jan 19 15:59:12 ccprod01 vmunix: scb->cdb: 12 01 80 00 ff 00
Jan 19 15:59:50 ccprod01 vmunix: scb->cdb: 12 00 00 00 80 00
Jan 19 15:59:55 ccprod01 vmunix: SCSI: Unexpected Disconnect -- lbolt: 481624, d
ev: cb016000, io_id: 10007b6'


While importing the volume group on the other cluster node I was getting the following message

'
Jan 19 15:56:37 ccdev01 vmunix: 20: CTEST6: 01 CTEST5: 01 C0
Jan 19 15:56:37 ccdev01 vmunix: 24: DCMDDBC: 0e000000
Jan 19 15:56:37 ccdev01 vmunix: 28: DNAD: 000803c5
Jan 19 15:56:37 ccdev01 vmunix: 2c: DSP: 00080048
Jan 19 15:56:37 ccdev01 vmunix: 30: DSPS: 000803c0
Jan 19 15:56:37 ccdev01 vmunix: 34: SCRATCHA: ff001000
Jan 19 15:56:37 ccdev01 vmunix: 38: DCNTL: 21 DWT: 00 D0
Jan 19 15:56:37 ccdev01 vmunix: 3c: ADDER: 00100408
Jan 19 15:56:37 ccdev01 vmunix: 40: SIST1: 00 SIST0: 40 Sf
Jan 19 15:56:37 ccdev01 vmunix: 44: GPCNTL: 0f MACNTL: 00 S1
Jan 19 15:56:37 ccdev01 vmunix: 48: RESPID1: 00 RESPID0: 40 Sc
Jan 19 15:56:37 ccdev01 vmunix: 4c: STEST3: 80 STEST2: 20 S7
Jan 19 15:56:37 ccdev01 vmunix: 50: RESV50: ff RESV51: ff S7
Jan 19 15:56:37 ccdev01 vmunix: 54: RESV54: ff RESV55: ff S1
Jan 19 15:56:37 ccdev01 vmunix: 58: RESV58: ff RESV59: ff S1
Jan 19 15:56:37 ccdev01 vmunix: 5c: SCRATCHB: 00000000
Jan 19 15:56:37 ccdev01 vmunix: Zalon register dump:
Jan 19 15:56:37 ccdev01 vmunix: IO_STATUS: f9000040
Jan 19 15:56:37 ccdev01 vmunix: IO_II_DATA: f9000040
Jan 19 15:56:37 ccdev01 vmunix: IO_DC_DATA: 80808080
Jan 19 15:56:39 ccdev01 vmunix:
Jan 19 15:56:39 ccdev01 vmunix: SCSI: Resetting SCSI -- lbolt: 25001020, bus: 1
Jan 19 15:56:39 ccdev01 vmunix: SCSI: Reset detected -- lbolt: 25001020, bus: 1
Jan 19 15:56:50 ccdev01 vmunix: 28: DNAD: 000803c5
Jan 19 15:56:50 ccdev01 vmunix: 2c: DSP: 00080048
Jan 19 15:56:50 ccdev01 vmunix: 30: DSPS: 000803c0
Jan 19 15:56:50 ccdev01 vmunix: 34: SCRATCHA: ff001000
Jan 19 15:56:50 ccdev01 vmunix: 38: DCNTL: 21 DWT: 00 D0
Jan 19 15:56:50 ccdev01 vmunix: 3c: ADDER: 00100408
Jan 19 15:56:50 ccdev01 vmunix: 40: SIST1: 00 SIST0: 40 Sf
Jan 19 15:56:50 ccdev01 vmunix: 44: GPCNTL: 0f MACNTL: 00 S1
Jan 19 15:56:50 ccdev01 vmunix: 48: RESPID1: 00 RESPID0: 40 Sc
Jan 19 15:56:50 ccdev01 vmunix: 4c: STEST3: 80 STEST2: 20 S7
Jan 19 15:56:50 ccdev01 vmunix: 50: RESV50: ff RESV51: ff S7
Jan 19 15:56:50 ccdev01 vmunix: 54: RESV54: ff RESV55: ff S1
Jan 19 15:56:50 ccdev01 vmunix: 58: RESV58: ff RESV59: ff S1
Jan 19 15:56:50 ccdev01 vmunix: 5c: SCRATCHB: 00000000
Jan 19 15:56:50 ccdev01 vmunix: Zalon register dump:
Jan 19 15:56:50 ccdev01 vmunix: IO_STATUS: f9000040
Jan 19 15:56:50 ccdev01 vmunix: IO_II_DATA: f9000040
Jan 19 15:56:50 ccdev01 vmunix: IO_DC_DATA: 80808080
Jan 19 15:56:52 ccdev01 vmunix:
Jan 19 15:56:52 ccdev01 vmunix: SCSI: Resetting SCSI -- lbolt: 25002320, bus: 1
Jan 19 15:56:52 ccdev01 vmunix: SCSI: Reset detected -- lbolt: 25002320, bus: 1
'

Any help please.

Thanks
John Jayaseelan

[Robert-Jan Goossens]

John,

Call your local HP crc, there could be problems with disk, controller, cable or terminator.

Robert-Jan.

[Rita C Workman]

There are different approaches taken by others to address your needs. So I can only say the way I prefer to do it...for myself when adding physical disk(s) to vg that is in an MC/SG environment...I take the package down! My reason:
The pkg is set to exclusive mode and I want to control of the vg..not the cluster. So I always:
cmhaltpkg pkgname
vgchange -c n /dev/vgname
vgchange -a y /dev/vgname
(Make my changes to vg)
..then reverse the above when done..
vgchange -a n /dev/vgname
vgchange -c y /dev/vgname
cmrunpkg pkgname

Now I notice a couple things..one that PE to small issue. Take a look at the thread you pointed to by Robert and it will answer that for you.
The other is the disk your trying to add....you got an lbolt error and the address on that message takes you exactly back to the disk you said your trying to add to use for a mirror copy. You have a hardware problem there you need to address first. Either that or use another disk.

Last your question about making any changes to your pkg script...If you are just adding more disk to an already existing (hence already referenced in your pkg-script) than NO you don't have do anything to it..or push out anything except...you will need to make a new vg-mapfile using the export command and send that to the other nodes in cluster and vg-import it on them so they will be updated. That is it ! (If you had created a new vg then yes..you would edit your pkg script, rcp copy of this script to other nodes and still do the vgexport/import for new vg to other nodes)

Hope that helps....
Rita

[Rita C Workman]

Made a too simple comment above..let me correct:

On OTHER nodes before you run vgimport of new vgmapfile

vgexport /dev/vgname
( Yes blow away what is there and then re-create it using the same unique minor number)

mkdir /dev/vgname
mknod /dev/vgname/group c 64 0x0_0000

Now do your vgimport command

OK...now I feel better !
Rita

[Rita C Workman]

* * * TIP * * *

MC/SG Environment..

It is very helpful to add some spare disk to your packages and let them sit there for when you need to increase space in the future. By adding disk to the vg, even though not allocated out yet, you save yourself work later. Now you can simply take advantage of OnlineJFS and distribute space as needed, without taking package down or having to redo your exports/imports....


*** This is just a suggestion I thought I'd throw out there - NO Points For This Please - I view it as a freebie ***