zhangrucong1982 created FLINK-6741:
--------------------------------------
Summary: In yarn cluster model with high available， the HDFS file
is not deleted when cluster is shot down
Key: FLINK-6741
URL: https://issues.apache.org/jira/browse/FLINK-6741
Project: Flink
Issue Type: Bug
Reporter: zhangrucong1982
Assignee: zhangrucong1982
The flink version of 1.3.0 rc2. I use yarn cluster with high available.
1、the mainly configuration is:
high-availability.zookeeper.storageDir: hdfs:///flink/recovery.
2、I use the command "./yarn-session.sh -n 2 -d" to start a cluster;
3、I use the command "./flink run ../example/streaming/WindowJoin.rar" to summit
a job;
4、I use the flink cancel command to cancel job;
5、 I use the “./yarn-session.sh -id xxxx” command to attach to the yarn cluster
and use stop command to shutdown the cluster.
6、After shutdown the cluster，I find the file in hdfs is not deleted. Like the
following:
/flink/recovery/application_1495781150990_0006/blob/cache/blob_9b2a6f6535819075889ebcf64490f4e6528b07
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)
