[FLINK-25468] Local recovery fails if local state storage and RocksDB working directory are not on the same volume

• Type: Bug
• Status: Closed
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 1.13.5, 1.14.2, 1.15.0
• Fix Version/s: 1.13.6, 1.14.3, 1.15.0
• Component/s: Runtime / State Backends
• Labels: pull-request-available
• Assignee: Till Rohrmann
• Reporter: Till Rohrmann
• Created: 28/Dec/21 14:26
• Updated: 31/Dec/21 06:21
• Resolved: 29/Dec/21 08:30

Description

Local recovery with RocksDB fails if the state storage directory is not on the same volume as RocksDB's working directory. The reason is that the RocksDBHandle only tries to hard link the RocksDB files when calling restoreInstanceDirectoryFromPath. If hard linking is not supported, then the operation fails.

In order to harden this behaviour, I suggest to fall back to copying the files over if hard linking fails.

Issue Links

• fixes: FLINK-10954 Hardlink from files of previous local stored state might cross devices
• links to: GitHub Pull Request #18222