02-29-2012 01:25 AM
Is there an easy way to move an existing cmu server setup to a new server?
I.e i currently run cmu on a proliant dl380 g5, and will retire this machine and move to a bl460c g7.
If somebody has some steps/hints on the best way of performing this move, it would be very appriciated.
Solved! Go to Solution.
03-05-2012 07:17 AM
Here are some general high-level steps to move the CMU server from one server to another. A complete set of instructions will vary between different cluster configurations. These steps focus on moving CMU, and will give you an idea of the process required to move the CMU server to a new node. Hopefully they will be helpful. Further assistance can be obtained by contacting CMU Support. If you are using CMU in a diskless environment, you will want to contact CMU Support for additional details.
Prepare the bl460c g7 for CMU installation as per the CMU documentation (install dependencies, configure tftp, nfs, and dhcp, configure networks, etc). Before installing the CMU RPM on the bl460c g7, shutdown the CMU, DHCP, and NFS services on the dl380 g5 and remove the CMU RPM (rpm -e cmu). Note that removing the CMU RPM will display a command to restore your configuration using the CMU restoreConfig command. Save this command, as it will be required later.
Copy the contents of /opt/cmu/ on the dl380 g5 to the bl460c g7. Be sure to preserve all soft-links and permissions.
Copy /etc/hosts and /etc/exports from the dl380 g5 to the bl460c g7. Update /etc/hosts with the new bl460c g7 IP address, and update /etc/exports as necessary. Restart NFS and test that you can mount one of the filesystems in /etc/exports from another node. Note that you may have to remove the 'fsid' field in each of the /etc/exports entries and restart NFS to get this to work.
Install the CMU RPM on the bl460c g7. When finished, run the restoreConfig command that was displayed when you removed the CMU RPM from the dl380 g5. Run this command on the bl460c g7.
Edit cmuserver.conf and update CMU_CLUSTER_IP to match the admin network IP of the bl460c g7. You may want to perform a search of /opt/cmu for all instances of the dl380 g5 IP address and update these entries with the new bl460c g7 IP address. For example, your kickstart files may have the dl380 g5 IP address configured as the NFS server.
Run the '/etc/init.d/cmu unset_audit' command.
Run the '/etc/init.d/cmu start' command to start the CMU service.
Verify that everything is working as expected. If necessary, use the dl380 g5 as a backup to compare configuration files, etc. Keep in mind that there may be configuration settings on the compute nodes that are specific to the dl380 g5, such as the NTP and NFS server configurations. You'll need to change these before retiring the dl380 g5.
Note that you can reinstall CMU on the dl380 g5 and run the restoreConfig command to restore this system back to its original configuration. However, please avoid running two instances of CMU on the same cluster at the same time, as this can confuse the monitoring and any provisioning. If necessary, you can shutdown CMU and DHCP on the bl460c g7 and then start them on the dl380 g5, and you will be back to your original configuration.
Hope this helps.