[Bug] [Zeta] Java process does not receive SIGTERM when running SeaTunnel inside container
#9827 opened on Sep 5, 2025
Description
Search before asking
- I had searched in the issues and found no similar issues.
What happened
When running SeaTunnel inside a Docker/Kubernetes container, the Java process cannot gracefully shut down because it never receives the SIGTERM signal sent by the container runtime. Currently, the startup script (bin/seatunnel-cluster.sh) launches Java like this (when not using -d):
java ${JAVA_OPTS} -cp ${CLASS_PATH} ${APP_MAIN} ${args}
In this case: • The shell script itself becomes PID 1 in the container. • The Java process runs as a child process of the shell. • When the container is stopped, SIGTERM is only delivered to the shell script (PID 1), not to the Java process. • The script exits immediately without forwarding the signal, leaving Java as an orphan process. • Finally, the Java process is killed by SIGKILL after the grace period, preventing SeaTunnel from shutting down gracefully.
This causes issues such as: • Shutdown hooks in Java never being executed. • Potential data loss or corruption when SeaTunnel tasks are abruptly killed.
SeaTunnel Version
2.3.11
SeaTunnel Config
none
Running Command
./bin/seatunnel-cluster -r xxx
Error Exception
none
Zeta or Flink or Spark Version
No response
Java or Scala Version
No response
Screenshots
No response
Are you willing to submit PR?
- Yes I am willing to submit a PR!
Code of Conduct
- I agree to follow this project's Code of Conduct