[ https://issues.apache.org/jira/browse/HDFS-14645?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16898025#comment-16898025 ]
Jinglun commented on HDFS-14645: -------------------------------- Hi [~jojochuang] [~Jihyun.Cho], I was thinking closing all children filesystems will break the semantics of FileSystem.newInstance(). Because the children filesystem are cached by FileSystem.Cache and might be shared by many ViewFileSystems. FileSystem.newInstance() means getting an independent FileSystem and closing it should not affect any other FileSystem object. Some software like Spark relys on the semantic. I think we should and add an inner cache caching all children FileSystems in ViewFileSystem like in HADOOP-15565. Let FileSystem.Cache caches ViewFileSystem only. So we can keep the semantic and the FileSytem.Cache makes sure there won't be too many FileSystem objects. > ViewFileSystem should close the child FileSystems in close() > ------------------------------------------------------------ > > Key: HDFS-14645 > URL: https://issues.apache.org/jira/browse/HDFS-14645 > Project: Hadoop HDFS > Issue Type: Bug > Components: hdfs > Affects Versions: 2.7.8, 3.3.0 > Reporter: Jihyun Cho > Assignee: Jihyun Cho > Priority: Major > Attachments: HDFS-14645.001.patch, HDFS-14645.002.patch, > HDFS-14645.003.patch > > > {{ViewFileSystem}} uses superclass's {{close}} in current implementation. > It removes from {{FileSystem.CACHE}} without closing the child FileSystems. > To close properly, when FileSystem is closing, its child FileSystems also > should be closed. -- This message was sent by Atlassian JIRA (v7.6.14#76016) --------------------------------------------------------------------- To unsubscribe, e-mail: hdfs-issues-unsubscr...@hadoop.apache.org For additional commands, e-mail: hdfs-issues-h...@hadoop.apache.org