| 1 | === The newer(est) plan: === |
| 2 | |
| 3 | Build a local, experiment-specific database from files generated by the containers system. Use this local database (living on a pnode or experiment wide configuration/Chef) to configure containers (and eventually physical) nodes. |
| 4 | |
| 5 | 1. look through and document files generated or existing in the containers system. Working document is here: http://containers.deterlab.net/wiki/ContainerFilesDoc. |
| 6 | |
| 7 | 2. determine what information is needed to properly configure the nodes to be good DETERizens. Rough cut is: user accounts (id, name, etc.), mounts (user home dirs, scratch, project dir), network interfaces (addresses, subnets, filters?, etc). For pnodes it'd be nice to have Chef (or whatever) also configure the container networks. (This is not unreasonable as it exists in python in the containers system now, but would require a rewrite in Ruby if we use Chef. If we use Puppet it might be less painful as Puppet speaks python.) |
| 8 | |
| 9 | 3. use info from 2 to write a script/module that creates the local database. |
| 10 | |
| 11 | 4. rewrite the existing Chef recipes in {{{/share/chef/chef-repo}}} to use the new schemas. They currently use the boss DETER schemas. |
| 12 | |