[
https://issues.apache.org/jira/browse/HBASE-25790?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17337967#comment-17337967
]
Viraj Jasani commented on HBASE-25790:
--------------------------------------
FYI [~apurtell], a new shell command that provides balancer rejection decisions
from in-memory ring buffer. Just wanted to check if you are fine including this
in 2.4.3 release.
> NamedQueue 'BalancerRejection' for recent history of balancer skipping
> ----------------------------------------------------------------------
>
> Key: HBASE-25790
> URL: https://issues.apache.org/jira/browse/HBASE-25790
> Project: HBase
> Issue Type: Sub-task
> Components: Balancer, master
> Reporter: Zhuoyue Huang
> Assignee: Zhuoyue Huang
> Priority: Major
> Fix For: 3.0.0-alpha-1, 2.5.0, 2.4.3
>
>
> [HBASE-24528|https://issues.apache.org/jira/browse/HBASE-24528] implements a
> NamedQueue that saved recent history of balancer decision details. Since we
> want to know exactly why the balancer is skipping to run, we need a new
> namedQueue called 'BalancerRejection' to hold the history of all details
> information why the balancer was rejected to run(skipping). LogEntry of
> 'BalancerRejection' should contain the information below:
> 1. Reason, to show why this time balancer was rejected
> Maybe look like:
>
> {code:java}
> (cost1*multiplier1+cost2*multiplier2+...+costn*multipliern)/sumMultiplier=0.025
> <= minCostNeedBalance(0.35)
> {code}
> 2. Detail cost factors and multipliers
> Maybe look like:
> {code:java}
> org.apache.hadoop.hbase.master.balancer.StochasticLoadBalancer$RegionCountSkewCostFunction
> cost:0.0 multiplier:500.0
> org.apache.hadoop.hbase.master.balancer.StochasticLoadBalancer$MoveCostFunction
> cost:0.0 multiplier:7.0
> org.apache.hadoop.hbase.master.balancer.StochasticLoadBalancer$RackLocalityCostFunction
> cost:1.0 multiplier:15.0
> {code}
>
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
