AFS Standards

These standards will help AFS failover work well and ensure our stability even if we lose an AFS server.

• Every volume that has an RO replica must have one on the same host as the RW volumes. Otherwise, if the other hosts go down, then there will be no available RO volumes, so the RO volumes will be unavailable.
• No *.user or *.mail volumes should have RO replicas, otherwise there may be confusion if a user somehow gets an RO volumes. Also, you would lose data if you converted an RO volume to an RW one.
• Where possible, paths should be on /afs/ugcs instead of /afs/.ugcs

AFS Disaster recovery

• RO volumes can be converted to RW ones with "vos con". Make sure that you aren't losing any data first. You will also have to "vos remsite" the old one.
• You will need to run "fs checkvolumes" after converting an RO -> RW so that the client can update its cache of volume locations.