Rejoining Rebuilt Node to Cluster

[JxD]

I have a server that is part of a cluster, and that I just rebuilt with a new install of DC-OE. I’ve given the rebuilt server the same hostname, IP address, etc. as it had before. I copied over the cluster config file and package directories. Then I ran cmrunnode and the rebuilt node came up in the cluster; so far so good.

However, when I try to start any packages on the rebuilt node, I get the following error:

cmgetpkgenv : Unable to get package configuration.
Either no cluster configuration file exists or package is not configured.

And in fact, from the command line, cmgetpkgenv will not retrieve the pkg config info on the rebuilt node. Does anyone have an idea how I can go about addressing this issue?

[SUDHAKAR_18]

hello ,

copy the cluster and package configuration files from the active node.

run cmcheck conf and check.

if everything is fine then run cmapplyconf and then run cmcrunnode and add the node to cluster. (mount points also need to take care, i mean vgexport and import)

[Rita C Workman]

On one of the existing nodes in the cluster - did you:

cmviewcl

Are those returning-node's packages there?
I'm guessing they are not there...

So simply go back to the returning node and do your cmcheckconf -P , and apply them when the cmcheck runs clean.

Then when done, I'd go back and make sure that my cluster is solid by running cmcheckconf on it and addressing any warnings/errors until clean and then apply it to. So that everything is neat and tidy.

Just a thought,
Rita

[Ishwar_1]

'cmgetpkgenv : Unable to get package configuration.
Either no cluster configuration file exists or package is not configured'
===========================================================

cmgetpkgenv: Modular package fails with the error: 'Unable to retrieve package attribute'.


Problem: In clusters with over 100 packages configured, when the cluster and packages start simultaneously, some packages may fail to start. Errors may appear in the log files of modular package log files, saying Unable to retrieve package attributes, and that the package failed to start. Messages such as the following appear in the package log file:


Feb 7 09:00:50 root@aialik.cup.hp.com master_control_script.sh[11121]:
ERROR: Unable to retrieve package attributes.
Feb 7 09:00:50 root@aialik.cup.hp.com master_control_script.sh[11121]:
ERROR: Function sg_source_pkg_env


The messages indicate that the cmgetpkgenv command had difficulty obtaining information about the package that is required for the package scripts to run. This does not mean that the package cannot be started after the other packages are stable.

Solution
========
Workaround: If you see this problem, you can issue a cmrunpkg on the packages that have not

[Stephen Doud]

On a functioning node, use cmgetconf to collect/recreate all cluster and package configuration files, and perform a comprehensive cmcheckconf on all of them. This will help verify that the re-installed node configuration matches the cluster binary file. If there are any problems, correct as needed.

cmviewconf will show whether the package exists and which nodes are failover nodes for the package. Insure the package directory has been copied to the updated server if the packages are legacy packages (per-package control scripts used).

The fact that the message includes the text "no cluster configuration file exists or ..." may be a clue that there is a hostname resolution issue on the installed node.
Serviceguard is particular about the configuration of certain system files for commands to run properly.

Pages 197-201 in the following manual describe how to properly configure the system to allow all Serviceguard commands to run on the local node:
"Managing Serviceguard Seventeenth Edition, First Reprint December 2009"
at http://bizsupport2.austin.hp.com/bc/docs/support/SupportManual/c02057890/c020578
90.pdf

Also note that /etc/services and /etc/inetd.conf should contain certain network services that Serviceguard requires:
hacl
identd (a.k.a. auth)
registrar

[JxD]

Even though it was a fresh install of 11.31.1009 on the rebuilt server, there were a couple of patches that were suggested by a patch analysis as well as the day one patches for HPVM 4.2.5. After I installed these patches and rebooted the server, I was able to start and stop the packages as before, no cmgetpkgenv error.

Thanks to all who replied, have assigned points!

[JxD]

Closing thread.