[HDDS-7101] EC: ReplicationManager - handle UNHEALTHY replicas

• Type: Sub-task
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Fix Version/s: 1.4.0
• Labels: pull-request-available
• Assignee: Siddhant Sangwan
• Reporter: Stephen O'Donnell
• Created: 05/Aug/22 20:41
• Updated: 05/May/23 15:23
• Resolved: 16/Nov/22 09:57

Description

If the DN scrubber marks a container as UNHEALTHY, the EC Replication Manager logic needs to be able to deal with it. It needs to remove the unhealthy one, and create a new copy using EC Reconstruction.

However we need to be careful to perhaps create the new replica before removing the unhealthy one. Also, if there are too many unhealthy, it would make the container unrecoverable, as there are not enough healthy replicas to try.

The LegacyReplicationManager handles this in the handleUnstableContainer method, but I wonder if we would be better handling it in the underReplicationHandler, as it is basically a form of under replication.

The logic to deal with it, is also going to be different from EC and Ratis.

Issue Links

• links to: GitHub Pull Request #3962