MCSG Config Dilemma

[Rob Saunders]

Where I previously had a 3 node cluster, (node1, node2, node3) I now have only 2 nodes (node2 and node3).

I cannot cmdeleteconf (even with the -f option) to permit the creation of a new configuration on node2/2 because the node1 does not exist and this prevents the cmdeleteconf from completing.

I guess there's some tidying up to be done somewhere, but where and how ?

TIA

Rob

[Sanjay_6]

Hi,

you can try this,

Do a cmgetcong to generate the ascii configuration file from your present cluster binary file. Then modify this file to exclude node1 and do a cmcheckconf. Then do a cmapplyconf to apply the new ascii configuration which exludes the node1.

Hope this helps.

regds

[Sanjay_6]

Hi Rob,

Try this link from hp docs site,

http://docs.hp.com/cgi-bin/fsearch/framedisplay?top=/hpux/onlinedocs/B3936-90065/B3936-90065_top.html&con=/hpux/onlinedocs/B3936-90065/00/00/51-con.html&toc=/hpux/onlinedocs/B3936-90065/00/00/51-toc.html&searchterms=delete%20cluster%20node%20down&queryid=20020307-063832

This link is from itrc,

http://us-support3.external.hp.com/cki/bin/doc.pl/sid=7259a1ba1b2b90d7f4/screen=ckiDisplayDocument?docId=200000059167468

Hope this helps.

regds

[fg_1]

Robert

You need to do the following:

1) edit /etc/cmcluster/cmclnodelist and ensure that the 2 nodes remaining are present.

2)cmquerycl -v -C /etc/cmcluster/cmclconf.ascii -n node1 -n node2

3) cmcheckconf -v -C /etc/cmcluster/cmclconf.ascii

4) cmapplyconf -v -C /etc/cmcluster/cmclconf.ascii

5) cmruncl -f -v

6) cmviewcl -v

This is all designed to regenerate your cluster using only the 2 existing nodes.

Anything further let us know.

Gl.

FG.

[melvyn burnard]

If all else fails,:
1)backup the /etc/cmcluster directories ;-}
2)halt the cluster completely
3)edit the ascii file to just refer to the two nodes and ENSURE you have a cluster lock disc coinfigured.
4)remove the current binary cmclconfig on each node
5)cmcheckconf
6)vgchange -c n on the vg's
7)vgchange -a y for all the vg's to be used (this ensures they get writen with the new cluster id)
8)cmapplyconf
9)vgchange -a n all vg's
10) cmruncl

[A. Clay Stephenson]

Also, if this is a permanent demotion to a two-node cluster, don't overlook the necessity of a lock disk.

[Tim D Fulford]

Rob

You only have two machines in a previous 3 node cluster. IF that node has been removed & say re commissioned then you probably only have one choice.
# rm /etc/cmcluster/cmclconfig
then redo a new cluster.ascii with the node removed then
# cmapplyconf -C /etc/cmcluster/cluster.ascii -P /etc/cmcluster//.ascii -P...

Regards

Tim