Hot to load Apache Flink cluster (not HA) when server boots up

Question

I've installed Apache Flink cluster with Job Manger and 2 Task Manager machines on Debian. Documentation says that bin/start-cluster.sh should be run to start the cluster, which works just fine. However, what is the best way to start the cluster when server boots up?

I can think of 3 options: use supervisord, use init.d script, make it a service

Suggestions?

We run it as a systemd service and it works well for us. The init.d vs systemd decision should be based on what your machines are already using. I assume it will probably be systemd on recent versions of Debian — Frank Lauterwald
Thanks @FrankLauterwald, this is exactly what I was leaning towards. — Ivan Naumenko

schmichri schmichri · Accepted Answer · 2016-12-12T09:39:50

For completion an example script running on Ubuntu 16.04. .

[Unit]
Description=Apache Flink
Requires=network.target remote-fs.target
After=network.target remote-fs.target


[Service]
Type=forking
User=flink
ExecStart=/opt/flink/bin/start-cluster.sh
ExecStop=/opt/flink/bin/stop-cluster.sh
Restart=always

[Install]
WantedBy=multi-user.target

Hot to load Apache Flink cluster (not HA) when server boots up

1 Answers