I’m in the middle of attempting to convert from homegrown workflow system to Dagster. When resuming runs, Dagster seems to generate a new

run_id

. The existing code expects that resumes maintain the same run id (which determines working directories and so on). Is there an id in Dagster that uniquely identifies a run across resumes? I see references to a `root_run_id`…

Hi Keith. When a failed run is resume / retried, it gets a new run id, but it becomes part of a run group, which has the originating run id as the

root_run_id

(as you have identified). If you keep retrying runs, each new run will have the last attempted run as a

parent_run_id

, while keeping the original run id as the root. This run group shows up in the right side panel on the Run view, showing the run lineage. I’m not sure if that will serve your purpose. There is not a run id that uniquely identifies the entire group, but I can get the set of runs from the run group based on the common root run id.

Thanks for explaining, @prha. So it sounds like that root run id should be what I want. I was getting the

run_id

from the Context and setting it in the environment for the called process. I don’t see

root_run_id

documented for Contexts (https://docs.dagster.io/_apidocs/execution#contexts) but I’ll experiment and see if it’s on there.

Yes, I think that’s right

Ok thanks. So if the

root_run_id

== the original

run_id

, then that should get me what I need. Experimenting.

I think that sounds reasonable… mind filing an issue?

Sure will do!