[ https://issues.apache.org/jira/browse/PHOENIX-4631?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16379674#comment-16379674 ]
Hui Zheng commented on PHOENIX-4631: ------------------------------------ phoenix-core/src/main/java/org/apache/phoenix/mapreduce/PhoenixInputFormat.java I think the connection should be cloesd after splits get generated here. {code:java} private List<InputSplit> generateSplits(final QueryPlan qplan, final List<KeyRange> splits, Configuration config) throws IOException { Preconditions.checkNotNull(qplan); Preconditions.checkNotNull(splits); // Get the RegionSizeCalculator try(org.apache.hadoop.hbase.client.Connection connection = HBaseFactoryProvider.getHConnectionFactory().createConnection(config)) { RegionLocator regionLocator = connection.getRegionLocator(TableName.valueOf(qplan .getTableRef().getTable().getPhysicalName().toString())); RegionSizeCalculator sizeCalculator = new RegionSizeCalculator(regionLocator, connection .getAdmin()); {code} > PhoenixInputFormat should close connection after generateSplits() > ----------------------------------------------------------------- > > Key: PHOENIX-4631 > URL: https://issues.apache.org/jira/browse/PHOENIX-4631 > Project: Phoenix > Issue Type: Bug > Affects Versions: 4.10.0 > Reporter: Hui Zheng > Priority: Major > > In our sparkstreaming usecase which loads a phoenix table as a Dataset, it > will leak zookeeper connection(3 connections per batch) and lead to OOM > exception in its driver process. -- This message was sent by Atlassian JIRA (v7.6.3#76005)