[KUDU-2680] Tombstones should be ignored in kudu fs update_dirs tablet failure check

• Type: Bug
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 1.9.0
• Fix Version/s: 1.9.0
• Component/s: CLI, fs
• Code Review: https://gerrit.cloudera.org/c/12313/
• Assignee: Andrew Wong
• Reporter: Adar Dembo
• Created: 29/Jan/19 23:48
• Updated: 04/Feb/19 04:21
• Resolved: 04/Feb/19 04:21

Description

If you suffer a disk failure, all affected tablets will be tombstoned upon rereplication. However, these tombstones still participate in the pre-flight check done by kudu fs update_dirs, which means that they'll prevent the removal of the failed disk (unless --force is used).

The invariant here is that a tombstoned tablet doesn't have a persistent data_dir_group in its superblock, so it'll never have a working in-memory data dir group, which means it'll always fail the pre-flight check.