Idea to add a last-update timeout to scripts similar to OpenCue timeout_llu

[mwiebe]

Many programs produce steady streams of update messages, while having highly variable total completion time. Timeouts for the whole command don't work well in this case, but a timeout for receiving any message at all from the program would reliably detect if it has stalled.

This idea is to add another timeout option to the <Action> template schema that plays this role. It would be something like:

logMessageTimeout: <posinteger> | <posintstring> # @optional @fmtstring @extension FEATURE_BUNDLE_2

Then a template could do this:

...
        onRun:
          command: "{{Task.File.Render}}"
          args:
            - /Users/myusername/blender_demo/3d/pavillon_barcelone_v1.2.blend
            - /Users/myusername/blender_demo/output_frames
            - "1"
            - "2"
          logMessageTimeout: "{{Param.LogMessageTimeout}}"
...

[epmog]

Love the idea, I think it makes a lot of sense. Some potential other things to think about. This would be able to work in conjunction with the other timeout options? Basically an action total timeout + the last update timeout. That's what I'm getting from the proposal of it being a different item.

Maybe just something that's more common in windows applications due to gui vs. terminal executable differences, but should there also be an option to watch a file for updates? ie. a log file, or theoretically even an output artifact/image? The idea being that the process is still performing work, even if we have less visibility into it outside of stdout. That would be more of a last-update rather than a last-message timeout

[Borislav-Bobo-Petrov]

I am rendering with redshiftCmdLine on Linux, my only insight is the log. I have experienced some rare cases where the renderer prints a specific message - "Preparing ray tracing hierarchy for meshes" - and then hangs forever and never proceeds with texture conversion and raytracing. I would love to have a way to define a condition "If you see this particular message in the log and and no new lines are printed to the log after N seconds, requeue the task".

A typical good log looks like this

2026/04/15 14:53:46-07:00 Preparing ray tracing hierarchy for meshes
2026/04/15 14:55:30-07:00 	Time to process 6 meshes: 1m:44s

I have seen a renderer hanging at the former line for 20+ minutes and never printing the latter, and I had to requeue the Task manually. Automating that case and any similar cases where the log progress dynamics are known would be great.