[HBASE-25666] Explain why balancer is skipping runs

• Type: Improvement
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Fix Version/s: 3.0.0-alpha-1, 2.5.0, 2.4.4
• Component/s: Balancer, master, UI
• Assignee: Zhuoyue Huang
• Reporter: Clara Xiong
• Created: 15/Mar/21 18:54
• Updated: 30/Aug/21 22:57
• Resolved: 04/Jun/21 22:27

Description

Balancer needs to find the balance of keeping cluster balanced and stable. There is a configuration minCostNeedBalance as the threshold of imbalance to try to balance. Since we use a single score to combine all the factors we consider, it is hard for operation to understand why balancer is "stuck". We should add to master-status UI to show balancer is skipping runs and explain all factors considered, such as the weight and cost of all cost functions.

 

Issue Links

• is a child of: HBASE-25697 StochasticBalancer improvement for large scale clusters
• is related to: HBASE-24528 Improve balancer decision observability
• relates to: HBASE-25697 StochasticBalancer improvement for large scale clusters
• relates to: HBASE-25906 UI of master-status to show recent history of balancer desicion

Sub-Tasks

• NamedQueue 'BalancerRejection' for recent history of balancer skipping: Resolved
• UI of master-status to show a recent history of that why balancer was rejected to run: Resolved