[HBASE-26309] Balancer tends to move regions to the server at the end of list

• Type: Improvement
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Fix Version/s: 2.5.0, 3.0.0-alpha-2, 2.4.9
• Component/s: Balancer
• Hadoop Flags: Reviewed
• Assignee: Clara Xiong
• Reporter: Clara Xiong
• Created: 30/Sep/21 03:11
• Updated: 09/Nov/21 02:32
• Resolved: 03/Nov/21 02:31

Description

When we have cohosted replicas, cohosted replica distribution has priority to region count skew. So we won't see region count  getting balanced until the replicas distributed. During this time, master UI shows a drift of regions to the server at the end of list and causes significant overload. This shows a bias of random region pick and needs to be addressed.

Issue Links

• is a child of: HBASE-25697 StochasticBalancer improvement for large scale clusters
• relates to: HBASE-26311 Balancer gets stuck in cohosted replica distribution
• relates to: HBASE-25625 StochasticBalancer CostFunctions needs a better way to evaluate region count distribution
• links to: GitHub Pull Request #3723
• links to: GitHub Pull Request #3811
• links to: GitHub Pull Request #3812

Sub-Tasks

• Repro balancer behavior during iterations: Resolved