This is a redo of the branch spark-app-id-as-container-name (from cf74924629b4db32c2a9d78abb89e1151037a5e6).
The idea is to allow using Spark application IDs as "container names" collected by gProfiler. Logically, a "container name" is just an identifier describing a process or a set of processes across (potentially) multiple nodes. For Docker/k8s/ecs etc, container names are logical. In Spark/YARN, application IDs or YARN application names can be used instead.
I turned ContainerNamesClient to a generic base ContainerNamesClientBase which is implemented by ContainerNamesClient (the old) and SparkContainerNamesClient (the new).
Description
This is a redo of the branch
spark-app-id-as-container-name
(from cf74924629b4db32c2a9d78abb89e1151037a5e6).The idea is to allow using Spark application IDs as "container names" collected by gProfiler. Logically, a "container name" is just an identifier describing a process or a set of processes across (potentially) multiple nodes. For Docker/k8s/ecs etc, container names are logical. In Spark/YARN, application IDs or YARN application names can be used instead.
I turned
ContainerNamesClient
to a generic baseContainerNamesClientBase
which is implemented byContainerNamesClient
(the old) andSparkContainerNamesClient
(the new).Related Issue
https://github.com/Granulate/gprofiler/issues/869
How Has This Been Tested?
T.B.D