Problem:
We want to understand the memory usage of our Cloud Dataflow job, preferably per worker/VM. We'd like to add this to Stackdriver, but it seems we can only add cpu usage and disk I/O metrics.
Question:
Is there a way to add Cloud Dataflow memory usage to Stackdriver?
As described here you can use the --experiments=enable_stackdriver_agent_metrics
flag to enable Agent metrics for your Dataflow job. This will include JVM metrics such as memory/usage
.