Elliott Clark created HBASE-13682: ------------------------------------- Summary: Compute HDFS locality in parallel. Key: HBASE-13682 URL: https://issues.apache.org/jira/browse/HBASE-13682 Project: HBase Issue Type: Bug Reporter: Elliott Clark
Right now when the balancer needs to know about region locality it asks the cache in serial. When the cache is empty or expired it goes to the NN. On larger clusters with lots of blocks this can be really slow. That means that balancer is un-usable while masters are being restarted. -- This message was sent by Atlassian JIRA (v6.3.4#6332)