On 05/28/2013 03:44 PM, Valerio Pachera wrote: > /sheep/sdb was used before so it contains data. If you are joining it into a running cluster, the old data will be moved to .stale directory, marking those stale objects with epoch when it left > How is sheepdog going to "react"? > Will it behave like md unplug and rebuild missing data on /sheep/<newdir>? Yeah, the newly joined sheep will recover the latest data from cluster transparently, nothing to worry at all. After recovery finishes, the stale objects will all be purged. Thanks, Yuan |