Pertsistent volume configuration recommendations


Are there parts of Druid that should be persisted through container restarts / cluster failure? If so, what are these parts?



Hey Kiefer,

ZooKeeper, metadata store, and deep storage (all not technically part of Druid, but things that Druid depends on) are the only things that really must be persisted through restarts.

It can help with restart times to persist historical node disks too, since if you don’t, each time there’s a failover it will need to redownload its dataset from deep storage.