Currently spotted with the grscicoll-cache Sparkapplication job. Look into the issue and propose a solution that fit all the spark Applications.
An idea based on the Yunikorn docs could be to create all the tasks-group for in the job pod (First pod started in the sparkapplication life cycle) and use reference in driver / executor pods. There should be a task-group for each type of Sparkapplication e.g. downloads, occurrence-table-build, maps-tiles.
Currently spotted with the grscicoll-cache Sparkapplication job. Look into the issue and propose a solution that fit all the spark Applications.
An idea based on the Yunikorn docs could be to create all the tasks-group for in the job pod (First pod started in the sparkapplication life cycle) and use reference in driver / executor pods. There should be a task-group for each type of Sparkapplication e.g. downloads, occurrence-table-build, maps-tiles.