What's missing / TODO

• Automated, fully failsafe, rollbacks

  • quite hard to implement without snapshots in the filesystem (lvm2 ?)

• Add checksums/gpg signatures

  • not too hard to do, but mostly useless if you don't make sure all your contrib rpms/debs are also signed -> lots of work

• crond watcher and restarter

  • Need to get around to that one day :)

• clientinfo reporting over HTTP and not just NFS

  • would be a little work and require a specialized CGI with http upload

• backup getupdates when you sync a bad getupdates

  • critical missing piece if you update getupdates and make a fatal mistake
  • watch /var/log/getupdates and retrieve and run shell script from a central server if the last update is too old (try 4 bytes of the IP, and then 3, 2, 1, and just failsafe)
  • Switch to YUM instead of apt-get on Red Hat?

Notes: