List view
Repairs and schedules should be more adaptive by providing the following features: - if segments fail due to timeout, then timeout should be extended exponentially (x2) over retries to allow them to pass - if a segment fails repeatedly a specific number of times, it should be marked as UNREPAIRABLE. Once there's no more segment in NOT_STARTED state and if some segments were marked as UNREPAIRABLE, the repair run should be marked as FAILED. - if timeouts occurred during a scheduled repair, the schedule should be updated as follows: - if more than 20% of segments got a timeout extension, the number of segments per node should be doubled - if less than 20% of segments got a timeout extension then the timeout should be extended to twice the longest segment duration
No due date•3/3 issues closed- No due date•51/51 issues closed
- No due date•1/1 issues closed
big features - new storage options
No due date•7/7 issues closed