Configuring Cluster Lock VG

[Steve Bonds]

I seem to be having a chicken-and-egg problem getting a new 2-node cluster set up.

Let me summarize my predicament:

cmcheckconf wants the cluster lock volume group to be available on all nodes.

cmcheckconf requires the cluster to be down to change FIRST_CLUSTER_LOCK_PV.

"vgchange -c y" only works if the cluster software is running.

cmapplyconf requires the lock volume group to be activated. (vgchange -a e vglock)

The cluster cannot be halted while a cluster aware volume group is activated

How can I create a configuration where cmapplyconf has the volume group available on all nodes, yet the cluster software is not running?

[Sanjay_6]

Hi Steve,

Refer to the thread below.

http://docs.hp.com/cgi-bin/fsearch/framedisplay?top=/hpux/onlinedocs/B3936-90045/B3936-90045_top.html&con=/hpux/onlinedocs/B3936-90045/00/00/33-con.html&toc=/hpux/onlinedocs/B3936-90045/00/00/33-toc.html&searchterms=cluster%20lock&queryid=20011008-105101

Hope this helps.

Thanks

[Steve Bonds]

The trick seems to be the order of volume group activation:

node 1:
+ create VG
+ activate VG (-a y)

node 2:
+ import VG
+ activate VG (-a y), ignore error

node 1:
+ deactivate VG
+ start single node cluster
+ change to cluster aware (-c y)
+ stop single node cluster
+ check config
+ apply config

In my prior attempt the lock VG was not enabled on the second node before the "vgchange -c y" was done. This really confused things.

[Sanjay_6]

Hi Steve,

This is how you have to configure SG.

Node 1
Create VG (vgcreate / vgextend / lvcreate / lvextend)
Deactivate VG (vgchange -a n /dev/vg_name)
export map (vgexport -p -s -m map_name /dev/vg_name)
rcp map to other node (rcp map_name node2:/tmp/map_name)

Node 2
Create VG directory and group file
Import VG from the map (vgimport -s -m /tmp/map_name /dev/vg_name)
Activate the VG (vgchange -a y /dev/vg_name)
Deactivate the VG (vgchnage -a n /dev/vg_name)

Node 1
Create ascii conf file for cluster (cmquerycl ....)
Edit ascii conf file and set the Cluster conf info
Check the conf file for errors (cmcheckconf ....)
If errors, rectify them.
If no errors.
Activate Cluster Lock VG (vgchange -a y /dev/vg_lock)
Distribute Cluster Binaries (cmapplycl ......)
Deactivate the Cluster Lock VG (vgchange -a n /dev/vg_lock)
Start the Cluster (cmruncl -v)

This should start the cluster on all the node(s) configured into the cluster configuration.

Here Node 1 is assumed to be the cluster configuration node.

Look at the thread below for more info.

http://docs.hp.com/cgi-bin/fsearch/framedisplay?top=/hpux/onlinedocs/B3936-90045/B3936-90045_top.html&con=/hpux/onlinedocs/B3936-90045/00/00/33-con.html&toc=/hpux/onlinedocs/B3936-90045/00/00/33-toc.html&searchterms=configuring%20cluster&queryid=20011008-113601

Hope this helps.

Thanks

[melvyn burnard]

Your assumption is not quite correct.
The nodes in the cluster have to have the VG configured on each node, by using vgexport -p -m mapfile on the node you create the VG, and then vgimport -m mapfile on the other node(s) in the cluster.
After first creating the VG, it is automatically activated on the first node, and that is ALL you need, is the VG activated on just one node.

This is all documented in the Managing MC/ServiceGuard manual available at :
http://docs.hp.com/hpux/ha