Mkexp restart workflow stops after first restart interval (ICON 2025.10 public)

Hello, I am running the public ICON 2025.10 atmosphere-only model on an HPC cluster using mkexp. I generated a simple Jablonowski–Williamson (JW) experiment with

INITIAL_DATE = 2000-01-01T00:00:00Z

FINAL_DATE = 2001-01-11T00:00:00Z

The generated namelist contains
checkpointtimeintval = P1Y
restarttimeintval = P1Y
The model behaves as expected:

  • It runs successfully from 2000-01-01 to 2001-01-01.
  • It writes the restart directory: jw5_restart_atm_20010101T000000Z.nc/
  • It writes finish.status = RESTART.
  • It exits normally without any errors.
    The restart files are present and appear to be complete. However, the experiment does not continue automatically to the final experiment date (2001-01-11). It stops after the first restart interval.
    My questions are:
  1. Is this the expected behaviour in the public ICON 2025.10 release?

  2. After finish.status = RESTART, is the continuation of the experiment supposed to be handled automatically by mkexp, or is an external workflow manager (or manual resubmission) expected?

  3. If automatic restart chaining is supported, is there an additional mkexp configuration or workflow option that needs to be enabled? Any guidance would be greatly appreciated.

    While investigating, I noticed that run_start explicitly resolves to .trigger = “” in the generated *.dump, so the restart-submission block in DEFAULT.run.tmpl is not rendered. Is this the intended design, and if so, what is the expected mechanism for invoking the logical run job after the initial segment?

    Thank you.

Hi Srihari, many mkexp test setups chose to disable restarting by setting `.trigger` to empty. During development, this allows re-running tests after each change without warnings like “experiment has already started”.

To enable proper restart behavior, you can add

[jobs]
  [[run_start]]
    .trigger = run

to your experiment config file.

hi!
thank you for the clarification. I ended up figuring this out as well but your explanation clears up a lot of other puzzling thoughts I had on some of the choices by the developers. now it’s clear. thank you