[FLINK-7783] Don't always remove checkpoints in ZooKeeperCompletedCheckpointStore#recover()

• Type: Sub-task
• Status: Closed
• Priority: Blocker
• Resolution: Fixed
• Affects Version/s: 1.3.2, 1.4.0
• Fix Version/s: 1.3.3, 1.4.0
• Component/s: Runtime / State Backends
• Assignee: Aljoscha Krettek
• Reporter: Aljoscha Krettek
• Created: 09/Oct/17 14:20
• Updated: 28/Apr/21 07:47
• Resolved: 24/Oct/17 12:21

Description

Currently, we always delete checkpoint handles if they (or the data from the DFS) cannot be read: https://github.com/apache/flink/blob/91a4b276171afb760bfff9ccf30593e648e91dfb/flink-runtime/src/main/java/org/apache/flink/runtime/checkpoint/ZooKeeperCompletedCheckpointStore.java#L180

This can lead to problems in case the DFS is temporarily not available, i.e. we could inadvertently
delete all checkpoints even though they are still valid.

A user reported this problem on the mailing list: https://lists.apache.org/thread.html/9dc9b719cf8449067ad01114fedb75d1beac7b4dff171acdcc24903d@%3Cuser.flink.apache.org%3E

Issue Links

• causes: FLINK-22502 DefaultCompletedCheckpointStore drops unrecoverable checkpoints silently
• links to: GitHub Pull Request #4863
• links to: GitHub Pull Request #4870
• links to: GitHub Pull Request #4879