"Shared Nothing" clusters
- OK, I want to add another machine as a backup
- Allow one machine to access the disk at a time
- "Failover" the disk connection when a machine fails
- Failover the filesystem mount point and the server processes on top of that as a single resource group
- Failover is still not instantaneous: it may take a while to bring the new service process back up
- Mounting the filesystem on the backup server is exactly like mounting it locally after a crash: ext2 plus journaling makes it feasible!
- Use RAID over NBD to simulate a shared disk?
No notes for slide 10