[HDDS-10446] Refactoring Node2ObjectsMap, Node2PipelineMap, Node2ContainerMap

• Type: Improvement
• Status: Resolved
• Priority: Minor
• Resolution: Implemented
• Affects Version/s: 1.4.0
• Fix Version/s: 1.5.0
• Component/s: SCM
• Labels: pull-request-available
• Assignee: Andrei
• Reporter: Andrei
• Created: 01/Mar/24 07:14
• Updated: 05/Mar/24 09:32
• Resolved: 05/Mar/24 09:32

Description

There is a base class `org.apache.hadoop.hdds.scm.node.states.Node2ObjectsMap for Node2PipelineMap and Node2ContainerMap.

It is redundant because of:
1. Node2ContainerMap is used only in tests
2. Node2PipelineMap doesn't  need inheritance at all

Moreover Node2PipelineMap contains ConcurrentHashMap<UUID, Set<PipelineID>> and 2 synchronized methods:

• void addPipeline(Pipeline pipeline) -> compute if absent for concurrent hash map
• void removePipeline(Pipeline pipeline) -> compute if present for concurrent hash map

I suppose synchronized is redundant here for these two methods.

Proposed changes:
1. Remove org.apache.hadoop.hdds.scm.node.states.Node2ObjectsMap class
2. Move Node2ContainerMap to tests
3. Don't use synchronized for Node2PipelineMap::addPipeline and Node2PipelineMap::removePipeline

GitHub discussions here

Issue Links

• links to: GitHub Pull Request #6303