Gary Helmling created HBASE-16958:
-------------------------------------
Summary: Balancer recomputes block distributions every time
balanceCluster() runs
Key: HBASE-16958
URL: https://issues.apache.org/jira/browse/HBASE-16958
Project: HBase
Issue Type: Bug
Components: Balancer
Reporter: Gary Helmling
Assignee: Gary Helmling
Fix For: 1.3.0
The change in HBASE-16570 modified the balancer to compute block distributions
in parallel with a pool of 5 threads. However, because it does this every time
Cluster is instantiated, it effectively bypasses the cache of block locations
added in HBASE-14473:
In the LoadBalancer.balanceCluster() implementations (in
StochasticLoadBalancer, SimpleLoadBalancer), we create a new Cluster instance.
In Cluster.<init>, we call registerRegion() on every HRegionInfo.
In registerRegion(), we do the following:
{code}
regionLocationFutures.set(regionIndex,
regionFinder.asyncGetBlockDistribution(region));
{code}
Then, back in Cluster.<init> we do a get() on each ListenableFuture in a loop.
So while we are doing the calls to get block locations in parallel with 5
threads, we're recomputing them every time balanceCluster() is called and not
taking advantage of the cache at all.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
