[FLINK-28102] Flink AkkaRpcSystemLoader fails when temporary directory is a symlink

• Type: Bug
• Status: Closed
• Priority: Critical
• Resolution: Fixed
• Affects Version/s: 1.16.0, 1.15.2
• Fix Version/s: 1.17.0, 1.16.1
• Component/s: Runtime / RPC
• Labels: pull-request-available
• Assignee: Weijie Guo
• Reporter: Prabhu Joseph
• Created: 17/Jun/22 04:37
• Updated: 22/Nov/22 09:44
• Resolved: 31/Oct/22 01:34

Description

Flink AkkaRpcSystemLoader fails when temporary directory is a symlink

Error Message:

Caused by: java.nio.file.FileAlreadyExistsException: /tmp
        at sun.nio.fs.UnixException.translateToIOException(UnixException.java:88) ~[?:1.8.0_332]
        at sun.nio.fs.UnixException.rethrowAsIOException(UnixException.java:102) ~[?:1.8.0_332]
        at sun.nio.fs.UnixException.rethrowAsIOException(UnixException.java:107) ~[?:1.8.0_332]
        at sun.nio.fs.UnixFileSystemProvider.createDirectory(UnixFileSystemProvider.java:384) ~[?:1.8.0_332]
        at java.nio.file.Files.createDirectory(Files.java:674) ~[?:1.8.0_332]
        at java.nio.file.Files.createAndCheckIsDirectory(Files.java:781) ~[?:1.8.0_332]
        at java.nio.file.Files.createDirectories(Files.java:727) ~[?:1.8.0_332]
        at org.apache.flink.runtime.rpc.akka.AkkaRpcSystemLoader.loadRpcSystem(AkkaRpcSystemLoader.java:58) ~[flink-dist-1.15.0.jar:1.15.0]
        at org.apache.flink.runtime.rpc.RpcSystem.load(RpcSystem.java:101) ~[flink-dist-1.15.0.jar:1.15.0]
        at org.apache.flink.runtime.taskexecutor.TaskManagerRunner.startTaskManagerRunnerServices(TaskManagerRunner.java:186) ~[flink-dist-1.15.0.jar:1.15.0]
        at org.apache.flink.runtime.taskexecutor.TaskManagerRunner.start(TaskManagerRunner.java:288) ~[flink-dist-1.15.0.jar:1.15.0]
        at org.apache.flink.runtime.taskexecutor.TaskManagerRunner.runTaskManager(TaskManagerRunner.java:481) ~[flink-dist-1.15.0.jar:1.15.0]

Repro:

1. /tmp is a symlink points to actual directory /mnt/tmp

[root@prabhuHost log]# ls -lrt /tmp
lrwxrwxrwx 1 root root 8 Jun 15 07:51 /tmp -> /mnt/tmp

2. Start Cluster
./bin/start-cluster.sh

Issue Links

• is duplicated by: FLINK-29698 Job Manager fails with FileAlreadyExistsException if java.io.tmpdir points to a symbolic link
• is related to: FLINK-29728 TablePlanner prevents Flink from starting is working directory is a symbolic link
• is related to: FLINK-30139 CodeGenLoader fails when temporary directory is a symlink
• relates to: FLINK-30143 Table Store fails when temporary directory is a symlink
• links to: GitHub Pull Request #21125