Yan Yan created FLINK-10526: ------------------------------- Summary: Hadoop FileSystem not initialized properly on Yarn Key: FLINK-10526 URL: https://issues.apache.org/jira/browse/FLINK-10526 Project: Flink Issue Type: Bug Components: YARN Affects Versions: 1.4.0 Reporter: Yan Yan
When we were bumping Flink from 1.3.2 to 1.4.0, we noticed the HadoopFileSystem no longer reads flinkConfig. This prevents users to pass in custom HDFS configuration (e.g. core-site.xml) via _fs.hdfs.hadoopconf_. Specially it is due to the cached _HadoopFsFactory_ is initialized with an dummy Configuration ([[code|https://github.com/apache/flink/blob/master/flink-core/src/main/java/org/apache/flink/core/fs/FileSystem.java#L387]|https://github.com/apache/flink/blob/master/flink-core/src/main/java/org/apache/flink/core/fs/FileSystem.java#L387]), which prevents future _flinkConfig_ getting passed in ([[code|https://github.com/apache/flink/blob/master/flink-filesystems/flink-hadoop-fs/src/main/java/org/apache/flink/runtime/fs/hdfs/HadoopFsFactory.java#L84]|https://github.com/apache/flink/blob/master/flink-filesystems/flink-hadoop-fs/src/main/java/org/apache/flink/runtime/fs/hdfs/HadoopFsFactory.java#L84]). We have verified by calling _FileSystem.initialize(flinkConfig)_ explicitly in _YarnApplicationMasterRunner.run_ solve the issue. -- This message was sent by Atlassian JIRA (v7.6.3#76005)