These scripts are DESTRUCTIVE and make some rather large assumptions.
I use these scripts to rapidly build and NUKE my lab environment for SUSE Enterprise Storage.
They are developed against version 5 currently
- The biggest and most dangerous assumption is that the OS is installed on a single drive and all others are fair game to be wiped out.
- DNS resolution is functioning properly
- These scripts will be run from the admin node, named salt
- You have created and distributed an ssh key for root to all nodes
To use these scripts, create 3 files
-
osdnodes.lst -- A list of the resolvable names of all OSD nodes
-
cluster.lst -- A list of all the nodes in the cluster (Monitors, gateways, admin, OSDs, etc)
-
/$PWD/policy.cfg -- the policy.cfg file you wish to use for deployments
Optionally, you may also create a drive_groups.yml
After creating the three files, simply run the buildit.sh script.
extract the perfcluster.tgz file and move the performancecluster directory to /root e.g. mv performancecluster /root/
- not necessary, optional and currently disabled in script
extract the correct dmb_kern_tune..tgz file and move the dmb_kern_tune directory /srv/salt/ modify to match your needs apply by using this command: salt '' state.apply