If I have a set of parallel jobs and 1 fails and the other succeeds. I try the resume execution/retry failed nodes, it triggers both the jobs again. Is there any setting in rundeck which can trigger only the failed job and not rerun the entire group again ? Is this a bug ?
The reason is that job reference steps is running on a single job, by design, rundeck consider that as a parent job exection and not individual executions (child jobs, the parallel jobs). If you want to avoid this, you will run these jobs individually (in a job calling each job using rd cli or Rundeck api) in a inline-script step.
In that way you can retry only the failed execution.
Now, to resume since a failed step, you can use the Job Resume Plugin (only for Rundeck Enterprise). The plugin allows to resume in the job failed step.