servicegard prob

[Chi Yi Lee]

Hi

I'm planning to add several lvol under a cluster that is already cover with servicegard.
Normally adding a lvol will need to edit fstab, but what and where is the file for servicegard?

thanks~

[V.Tamilvanan]

Hi,

In service guard the enties will be in package control file . normally in /etc/cmcluster// pkgname.ctl.

For more details you go through the documentation at this site.

http://docs.hp.com/hpux/onlinedocs/B3936-90024/B3936-90024.html

[Jose Mosquera]

Hi,

You will done this task without stop your package. Create your news LV's, assigns their corresponding FS and mount them, all of them following standards procedures.

Normally you must have the SG control script placed into /etc/cmcluster//pkg_name.cntl.
Inside this file look for FILESYSTEMS section.

Specify the filesystems which are used by this package. Fill in the name of your first
logical volume, filesystem and mount option for the file system. You must begin with LV[0], FS[0] and FS_MOUNT_OPT[0] and increment the list in sequence. Please have special care in the sequence!!!

If you want, you can test your configuration or assume the risk of waiting the next activation of the package.

To test the package:
1.- Disable switching
#cmmodpkg -d -n
2.- Check switching status
#cmviewcl -v
3.- Halt your package
#cmhaltpkg
4.- Umount manually all new FS defined
5.- Prepare to check the startup secuence status into the package log
#tail -f /etc/cmcluster//pkg_name.cntl.log
6.- Start up you package
#cmrunpkg
7.- Analize previous log and bdf

BR.

[Rita C Workman]

...just remember if all your adding is lvol you can do this with the package up and make use of OnlineJFS.

But if your also adding disk for those new lvols, the package will need to be down.

And if your adding a new volume group don't forget to add the entry to your pkg.ascii file then you'll need to cmcheckconf -k -v -P and then cmapplyconf -k -v -P

Rgrds,
Rita

[Gary L. Paveza, Jr.]

Also, make sure that you have an updated volume group on each node that will run that package.

First, modify the volume group on the primary node, then do the following:

On the primary node create a volume group map file :

vgexport -m -s -p -v

NOTE: Make sure you use the -p option - that leaves the volume group intact and only creates the map file.

On any system which runs that node (it should have that volume group defined).

- Copy the mapfile created above to this node
- Do a ll in /dev/ and make a note of the minor number assigned to the volume group
- vgexport to get rid of the old definition
- mkdir /dev/
- mknod /dev//group c 64 0x0000
- vgimport -m -s -v

[Stephen Doud]

This and many other "How to" documents for Serviceguard can be found in the ITRC knowledge database.

This particular topic is discussed in the document:

UXSGLVKBAN00000008
TITLE: Adding Logical Volumes to a ServiceGuard volume group

-SD-

[Chi Yi Lee]

Hi

thanks for the responses.
I edited package.sh and add lvol and so for.
Now I have a new problem.
From Server A, I am able to halt and run package without any problem, but I am not able to do the other way.

Server A has vg_fods1
vg_fods1 can switch to server B

I edited the shell in Server B under vg_fods1 (package name) just like in Server A.

When I apply cmrunpkg -n serverB vg_fods1

error occured:
vxfs fsck: /dev/vg_fods1/rlvol_store:cannot stat /dev/vg_fods1/rlvol_store
ERROR: Function check_and_mount
ERROR: Failed to fsck one of the logical volumes.

why did this happen since it worked in serverA?

thanks

[Jonas Linde]

It's because you need to vgepxort the VG on the node where you added the lvol, and vgimport it on all other nodes. You are missing the device file for your new lvol.

Adding lvol and vgexport/import can be done without stopping the package.

Jonas

[Rita C Workman]

Follow what Jonas said about vgexport/vgimport...

But just in case you did that already..and still are seeing an error...

I have found it best to NOT edit the other pkg control script. But rather, once the pkg control script is fixed and working on the primary node, I rcp this file to the other node(s). That way I don't make a 'fat finger' error and wonder why it won't work on the failover node.

Just a thought,
Rita

[Bharat Katkar]

Hi,
Have a look at this doc.

http://docs.hp.com/hpux/onlinedocs/B3936-90073/B3936-90073.html

Hope that helps.
Regards,