I'm not so sure. The problem is much more complex. This is a problem of managing configuration files, updating system, installing new software and testing. All this problems are connected. - All services should be clustered if it is possible - We need an ability to add new node to cluster fast /delete old node from cluster transparently - We need an ability to update base OS version for a cluster - We need an ability to have custom configs on each node and to propagate changes to each node in a cluster (e.g. after base OS update changes in configs should be propagated to each node) - We need an ability to create alternative test environment and update it/install new soft ther, test and switch to it after testing without loosing applications data - We need an ability to restore each node to its previous state - We need to know who of admins, when and why changed configs
We investigate the following approach now: 1) The root fs of each server is not persistent (doesn't store changes), so to create new node we need to connect new blade to this disk and make customizations in /etc, may be /usr/local/etc or some other places. /etc, /usr/local/etc mounted rw from separate place 2) Now we need an ability to propagate changes to /etc. SVN merge seems a good candidate... So we may quickly clone a disk (e.g. on storage system), update system (or soft) and create test environment. Configs in svn allow you to rollback and view changes to them... However, we are only in early stages of planning and investigate different approaches. > This sounds like > something CFengine would be a good solution > for. -- This message posted from opensolaris.org _______________________________________________ opensolaris-discuss mailing list opensolaris-discuss@opensolaris.org