When a drive becomes unavailable, the S Series Node immediately starts repairing the objects with data or parity chunks on that drive by re-creating those chunks on other drives. A chunk for an object can be re-created as long as both of these are true:
- At least 20 other chunks for that object are on available drives.
- At least one other drive:
- Is available
- Doesn't already have a chunk for that object
- Has enough space to store the re-created chunk
Because the S Series Node starts repairing objects immediately, the chance of more than six chunks for an object being unavailable at the same time is exceedingly low. Also, when making repairs, the S Series Node gives higher priority to objects that have a higher risk of becoming irreparable.