[ https://issues.apache.org/jira/browse/FLINK-10346?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16620027#comment-16620027 ]
Paul Lin commented on FLINK-10346: ---------------------------------- It seems that the same problem exists in FsStateBackend too. The checkpoint directories and HA directories are not cleaned up after the cluster shuts down, and Zookeeper is also flooded with per cluster nodes (something like /flink/${applicationId}) in case of per job clusters. > MemoryStateBackend does not clean up checkpoint directory > --------------------------------------------------------- > > Key: FLINK-10346 > URL: https://issues.apache.org/jira/browse/FLINK-10346 > Project: Flink > Issue Type: Bug > Components: State Backends, Checkpointing > Affects Versions: 1.5.3, 1.6.0, 1.7.0 > Reporter: Till Rohrmann > Priority: Major > Fix For: 1.7.0 > > > The {{StateBackendLoader}} creates a random subdirectory under > {{HighAvailabilityOptions.HA_STORAGE_PATH}} for the {{MemoryStateBackend}} if > no checkpointing directory has been specified (see > {{StateBackendLoader.java:246}}). The problem is that this directory gets > never removed after the {{MemoryStateBackend}} is closed. It is important to > clean up these created directories. -- This message was sent by Atlassian JIRA (v7.6.3#76005)