Search code examples
monitoringapache-flinkprometheus

How could I override configuration value in Apache Flink?

I'm trying to gather metrics from Apache Flink into Prometheus. Flink documentation says that I need to add following lines to my flink-conf.yaml:

metrics.reporter.promgateway.class: org.apache.flink.metrics.prometheus.PrometheusPushGatewayReporter
metrics.reporter.promgateway.host: localhost
metrics.reporter.promgateway.port: 9091
metrics.reporter.promgateway.jobName: myJob

I want to mark different jobs with different names inside of Prometheus. How could I override configuration parameter metrics.reporter.promgateway.jobName on per-job basis (each job is running inside of its own Flink cluster session)?

There is a couple of problems:

  • I can't override flink-conf.yaml. I've found only FLINK_CONF_DIR parameter to override whole configuration directory. But it doesn't look like a right solution to override configuration directory for every single job.
  • I can't override initial configuration of StreamExecutionEnvironment because it is being constructed inside of StreamExecutionEnvironment.getExecutionEnvironment method and can't be modified after environment's initialization.
Solution

  • You can modify the effective configuration by specifying a dynamic property when starting a Flink job cluster. Assuming that you are deploying to Yarn the command would look like:

    bin/flink run -m yarn-cluster -yD metrics.reporter.promgateway.jobName=myCustomJob <USER_CODE_JAR>

    The dynamic properties are sent to the Yarn cluster and overwrite existing configuration key value pairs.