[HBASE-21338] [balancer] If balancer is an ill-fit for cluster size, it gives little indication

• Type: Sub-task
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Fix Version/s: 3.0.0-alpha-1, 2.2.0, 2.1.1, 2.0.3
• Component/s: Balancer, Operability
• Labels: balancer
• Hadoop Flags: Reviewed

Release Note

The description claims the balancer not dynamically configurable but this is an error; it is http://hbase.apache.org/book.html#dyn_config

Also, if balancer is seen to be cutting out too soon, try setting "hbase.master.balancer.stochastic.runMaxSteps" to true.

Adds cleaner logging around balancer start.

• Assignee: Xu Cang
• Reporter: Michael Stack
• Created: 18/Oct/18 16:36
• Updated: 22/Apr/19 23:45
• Resolved: 23/Oct/18 23:30

Description

See parent issue. Running balancer on a cluster where the max steps was way inadequate, the balancer gave little to no indication that it was ill-configured. In fact, it only logged its starting and then that there was nothing to do though the cluster was obviously out-of-whack.

Ideally the balancer would complain when say the maxSteps limit is a small fraction of what the cluster's calculated max steps are, or it would notice that the balancer is making little progress on an imbalanced cluster and shout. Can we set balancer configs w/o having to restart Master?

Issue Links

• is related to: HBASE-21373 Backport to branch-1, "HBASE-21338 [balancer] If balancer is an ill-fit for cluster size, it gives little indication"