MC/SG Mirroring

[kholikt]

I have two servers which is running 11.23 IA platform. Both servers are connected to MSA 30. Currently the shared volume is using mirroring to provide redundancy. I only have experienced for MC/SG in VA setup.

Just wondering what is the step if the following scenario happen.

1. One of the mirrored disk failed, a new disk is replaced. Can I do without shutdown the cluster nodes.

2. Adding additional new disk and setup the mirroring.

3. In the case that if the cluster need some troubleshooting. Can I deactivate the VG and mount it to one server only without startup of cluster package?

4. Adding additional volume group to the cluster node

[A. Clay Stephenson]

1) Yes, the procedure is the same. Of course, the disk replacement needs to be done on the currently active node.

2) If you are adding new disks (PV's) to an existing VG then you do the add on the currently active node. You will then need to do a vgexport and vgimport the revised VG on each adoptive node.

3) Yes. You should note that the VG is now cluster aware so the cluster itself should remain up although the particular package can be down.

4) If this is a shared VG then you must import the VG on all nodes; if not then no additional work is required.

All of this is covered in the MC/SG manuals.

[Stephen Doud]

1. One of the mirrored disk failed, a new disk is replaced. Can I do without shutdown the cluster nodes.

If patches PHKL_32095 and PHCO_31709 (or later versions) are installed, a new option (-a n/y) is added to pvchange to detach and
re-attach a PV to the volume group. This option can be used to replace a mirror disk without interrupting the operation of a
package.
Whitepaper discussing it:
http://docs.hp.com/en/7161/LVM_OLR_whitepaper.pdf

2. Adding additional new disk and setup the mirroring.
Assuming you are using LVM, this document describes how to add a disk to a
Serviceguard volume group: UXSGLVKBAN00000002
TITLE: Adding A Disk To A Volume Group In A ServiceGuard Package


3. Can I deactivate the VG and mount it to one server only without startup of cluster package?

When a cluster is down, a VG can be de-clustered and activated normally.
# vgchange -c n vg??
# vgchange -a y vg??

To get the VG re-clustered, use cmapplyconf on the cluster configuration file.

If the cluster is functional but you need to work on the LVM application
data, consider editting the package control script, commenting out the
customer_defined_run_cmds and customer_defined_halt_cmds and running the
package control script manually: start (use "stop" to halt it)

4. Adding additional volume group to the cluster node:
- Create the VG and it's logical volumes on one node
- VGimport the VG on the second node (see document in item 2 above)
- Edit the cluster configuration file, add a VOLUME_GROUP reference for the new VG
- either CMAPPLYCONF the file, or manually "clustered" the VG if the node is running cmcld:
# vgchange -c y vg??
- Edit the package control script, adding the new VG and Lvol references (remember to increment the VG[?] and LV[?] values)
- Copy the script change to adoptive nodes
- Test the updated script.