Error message: cmcld: Lookup-node-by-name failed.

[Robert Hunke]

Operating system that we are using is HPUX 11.11 with ServiceGuard OPS edition, version A.11.13.

When I try to bring up a newly created two node cluster, I get this error from cmcld: Lookup-node-by-name failed. We have verified remsh, rlogin from both nodes. We have verified services inetd.conf resolv.conf nssitch.conf /etc/host
cmnodelist .rhosts etc.
The cmapplyconf output is fine with no error indications. The binary file is properly distributed.

Have no idea how to solve this error . . . HELP!!!:)

[Jeff Schussele]

Hi Robert,

Don't know if your question contained a typo but the node list file should be named
cmclnodelist

HTH,
Jeff

[Robert Hunke]

Just a typo. I meant to type in my question cmclnodelist, it is correctly spelled on the cluster. Thanks!

[Dirk Wiedemann]

Hello Robert,

maybe its a DNS problem.
check on both systems lookup and reverse lookup with an without context.

Dirk

[Sajid_1]

Do you have this patch installed - PHSS_27087. If not install it, that may solve the issue.

[Jeff Schussele]

Hi (again) Robert,

It's a good idea to make sure that your /etc/hosts file contains entries for the node names used - even if as alias' to the actual or FQDN names. Then set nsswitch.conf to search files first. This will insure proper cluster ops even if DNS is down.
I even take it a step further & define names for the heartbeat IPs and get them in hosts as well as DNS.
Sure sounds like name resolution trouble.

Rgds,
Jeff

[Robert Hunke]

I checked the dns and it resolves both ways... Thanks

I did not have PHSS_27087 and I'll load that as soon as it comes down. Thanks!

[Robert Hunke]

Jeff,

I have the FQDN and the "short" name in the hosts list and I had modified the nsswitch.conf file to "file" first then dns.

I had also modified the entry to include
files[NOTFOUND=continue] dns in the nsswitch.conf

[Robert Hunke]

I have applied the patch PHSS_27087 with negative results :-(

I have rechecked dns and can resolve FQDN, alias and ips. All seem ok.

[Sajid_1]

Robert,

Just another try - check each system for the /.rhosts file. If you don't have entries for the server and root there, add it.

gl,

[Robert Hunke]

Sajid,
I didn't quite understand what you meant by /.rhosts. My root home account is under /home/root.
I have verified my .rhosts file to have the proper enteries. I have also verified the cmclnodelist. Does anyone know where the cmcld daemon is looking for the node name? I thought it is defined in the cmclnodelist file.

[Jeff Schussele]

Robert,

Make sure you ONLY use the shortname for the NODE_NAME in the cluster configuration file.
Using the FQDN name in that file can cause the problem you're seeing.

See the following thread:

http://forums.itrc.hp.com/cm/QuestionAnswer/1,,0xc436cbaac6dcd5118ff40090279cd0f9,00.html

Also you may want to verify that there are no unprintable chars in the config file, cmclnodelist or .rhosts files - including spaces. vi those files & double-check for leading white-space in the .rhosts & cmclnodelist files & for trailing white-space in the config file.

Jeff

Rgds,
Jeff

[Robert Hunke]

Jeff,

It was the hostname!!! Thank you for your suggestion and the very helpful link. The cluster is up an running:)!!!

Thanks to all for your help and all of the great suggestions!

Regards,
Bob and Deb