Sysadmin:Disaster

We need to set up a disaster recovery plan. Disaster is something that breaks the entire cluster- either a crucial machine goes down, or Kerberos or AFS gets hacked.

Kerberos/AFS gets hacked

 * What to do: 1) Panic 2) Pull the network cable 3) ??? 4) Profit!

Kerberos

 * New accounts would be suspect, but old passwords are hashed
 * Keep a copy of the kerberos database on a secure CD

AFS

 * Restore from known good backups
 * Change localauth keys on all three servers