Setting up a cluster from scratch

[S.Rider]

I'm trying to setup a couple new servers in a cluster. When I run the "cmquerycl" command to get a config.ascii file to modify, I get the following error. Or is it just as easy to grab a config file from another cluster we have and modify that instead.

# cmquerycl -v -n ussbcsv1 -n ussbcsv2 -C cyclonedev.ascii

Begin checking the nodes...
Looking for other clusters ... Done
Gathering configuration information ..
Gathering storage information ..
.....
Gathering Network Configuration ..... Done
Error: Unable to connect to node ussbcsv1: Not owner
Error: Unable to connect to node ussbcsv2: Not owner
Warning: Not probing node ussbcsv1 as it is currently unreachable.
This may cause network partitions to be reported.
Warning: Not probing node ussbcsv2 as it is currently unreachable.
This may cause network partitions to be reported.
Failed to gather configuration information.
ussbcsv1:[/etc/cmcluster]

[Denver Osborn]

did you setup /etc/cmclnodelist or rhosts?

http://forums1.itrc.hp.com/service/forums/questionanswer.do?threadId=803313

-denver

[Denver Osborn]

also, if you could post the ver of mc/sg it would help... later versions use identd by default. You can setup identd or disable it and use cmclnodelist or rhosts, your choice.

search the forums for something like "unable to connect to node not owner" and go over the mc/sg docs for the ver you're working with at http://docs.hp.com

http://forums1.itrc.hp.com/service/forums/questionanswer.do?threadId=865029

hope this helps,
-denver

[S.Rider]

I setup /.rhosts & /etc/cmcluster/cmclnodelist on both servers - it's still dieing.
ServiceGuard is A.11.16.00
Servers are both 11.11

Thanks for the help so far.

[Denver Osborn]

If cmquerycl doesn't look like it's using ident, then it could be your rhosts isn't setup right on the nodes.

ssh to one of the nodes and do 'who -muR' and compare the hostname with what's in your rhosts file. If they're not the same this could be a problem. you could use the ip address in the rhosts -vs- hostnames if there's a diff between the rhosts entry and name resolution.

Can you also see what's logged in the syslog.log when it fails.

-denver

[Denver Osborn]

Jay,

to speed things along... can you post the cmclnodelist, /etc/hosts entries for the nodes and verify /etc/nsswitch.conf points to files first.

/etc/hosts should have "ip_addr host host.fqdn" and cmclnodelist would have

node1 root
node2 root

on both nodes.

any details from syslog.log would help troubleshoot to. Chances are it could be name resoultion prob and that's why you see "Not owner" for both nodes when running the cmquerycl... or an identd related config prob could be slowing you down.

[melvyn burnard]

I would suggest somebasic tests first.
check hostname/ip lookup BOTH ways using nslookup, checking that you get the ip you expect for the hostname, and vice versa.
Ensure you have all th econfigured lan card ips of each node in your /etc/hosts file across all cluster nodes, and have /etc/nsswitch.conf pointing at files first, and then DNS (if used).
Also when using cmclnodelist, have all your possible hostname/ip addresses for the cluster nodes included, and if using ip addresses, make sure SG is patched as 11.16 had an issue that it would only work with hostnames in the cmclnodelist file.
Finally, make sure you read through the 11.16 manuals, and the security manuals/white papers available at http://docs.hp.com/en/ha under Serviceguard section