[HDDS-9652] Legacy RM will not replicate all unhealthy containers when some are decommissioning

• Type: Improvement
• Status: Resolved
• Priority: Major
• Resolution: Implemented
• Fix Version/s: 1.4.0
• Component/s: Ozone Datanode, SCM
• Labels: pull-request-available
• Assignee: Stephen O'Donnell
• Reporter: Stephen O'Donnell
• Created: 08/Nov/23 16:10
• Updated: 10/Nov/23 12:58
• Resolved: 10/Nov/23 12:58

Description

In a situation where a container is closed or quasi-closed with all unhealthy replicas, the legacy Replication Manager will not replicate the unhealthy containers if the under replication is caused by decommission.

Ie, if we have 3 replicas:

IN_SERVICE, unhealthy
IN_SERVICE, unhealthy
DECOMMISSIOING, unhealthy

Then the legacy RM will take no action and decommission will be stuck.

The new RM handles this state correctly and does not need to be changed.

Additionally, in KeyValueContainer.java, unhealthy replicas cannot be exported / replicated, so even if the RM was not present, replication still could not happen.

This change corrects both issues and adds a couple of extra tests to the new RM to prove it works ok.

Issue Links

• links to: GitHub Pull Request #5562