How to properly set up a monitoring script?

[zakkak]

I am trying to start an async script that will monitor the resource utilization while my benchmark script runs. To achieve this I tried starting the stats-printer script with a repeat-until command that instructs the code to run till the benchhmark script signals completion.

However, when running this scenario I am getting:

...
13:51:27,152 [ test:benchmark-275 @ server ] for-each: server = {"container":"dummy-maestro-instance","port":8080,"type":"dummy"}
13:51:27.510 queueing
  host=localhost
  script=stats-printer
13:51:27,511 [ test:benchmark-275 @ server ] script-cmd: stats-printer
13:51:27,511 [ :stats-printer-295 @ server ] stats-printer
13:51:27.511 signal dummy-bmark-done missing latch, defaulting to 0
13:51:27,512 [ :stats-printer-295 @ server ] queue-download: ${{WORKDIR}}/${{type}}-stats.txt=/tmp/qdup-maestro-test/dummy-stats.txt ./20250919_135114/server/.
13:51:32,917 [ test:benchmark-275 @ server ] sleep 5s
13:51:32.917 signal dummy-bmark-done missing latch, ignoring
13:51:32,917 [ :- @  ] reached dummy-bmark-done
...

The relevant qDup part looks like this:

    stats-printer:
        - repeat-until: ${{type}}-bmark-done
          then:
              - sleep: 1s #always include the sleep!
              - sh: ${{CONTAINER_RUNTIME}} stats --no-stream --format "{{.Name}} {{.MemUsage}}" ${{container}} >> ${{WORKDIR}}/${{type}}-stats.txt
        - queue-download: ${{WORKDIR}}/${{type}}-stats.txt .

    benchmark:
        - wait-for: maestro-native-controller-started
        - wait-for: maestro-jvm-controller-started
        - wait-for: dummy-maestro-ready
        - sh: cd ${{WORKDIR}}
        - for-each:
              name: server
              input:
                  - container: dummy-maestro-instance
                    port: 8080
                    type: dummy
                  - container: maestro-native-controller
                    port: 9500
                    type: native
                  - container: maestro-jvm-controller
                    port: 9600
                    type: jvm
          then:
              - script:
                  name: stats-printer
                  async: true
                with:
                    container: ${{server.container}}
                    type: ${{server.type}}
              - sh: sleep 5s # emulation of actually doing something
              - signal: ${{server.type}}-bmark-done
              - sh: mv ${{WORKDIR}}/run/0000 ${{WORKDIR}}/${{server.type}}-run || echo "No detailed report"
              - sh: rm -rf ${{WORKDIR}}/run
              - sh: ${{CONTAINER_RUNTIME}} stats --no-stream --format "{{.Name}} {{.MemUsage}}" ${{server.container}}
              - queue-download: ${{WORKDIR}}/${{server.type}}-run .
              - queue-download: ${{WORKDIR}}/report-${{server.type}}.html .
        - signal: bmark-done

[zakkak]

It looks like the proper way to achieve this is by using set-signal, e.g.:

    setup-printer-signals:
        - set-signal: dummy-bmark-done
        - set-signal: native-bmark-done
        - set-signal: jvm-bmark-done

    stats-printer:
        - repeat-until: ${{type}}-bmark-done
          then:
              - sleep: 1s #always include the sleep!
              - sh: ${{CONTAINER_RUNTIME}} stats --no-stream --format "{{.Name}} {{.MemUsage}}" ${{container}} >> ${{WORKDIR}}/${{type}}-stats.txt
        - queue-download: ${{WORKDIR}}/${{type}}-stats.txt .

    benchmark:
        - wait-for: maestro-native-controller-started
        - wait-for: maestro-jvm-controller-started
        - wait-for: dummy-maestro-ready
        - sh: cd ${{WORKDIR}}
        - for-each:
              name: server
              input:
                  - container: dummy-maestro-instance
                    port: 8080
                    type: dummy
                  - container: maestro-native-controller
                    port: 9500
                    type: native
                  - container: maestro-jvm-controller
                    port: 9600
                    type: jvm
          then:
              - script:
                  name: stats-printer
                  async: true
                with:
                    container: ${{server.container}}
                    type: ${{server.type}}
              - sh: sleep 5s # emulation of actually doing something
              - signal: ${{server.type}}-bmark-done
              - sh: mv ${{WORKDIR}}/run/0000 ${{WORKDIR}}/${{server.type}}-run || echo "No detailed report"
              - sh: rm -rf ${{WORKDIR}}/run
              - sh: ${{CONTAINER_RUNTIME}} stats --no-stream --format "{{.Name}} {{.MemUsage}}" ${{server.container}}
              - queue-download: ${{WORKDIR}}/${{server.type}}-run .
              - queue-download: ${{WORKDIR}}/report-${{server.type}}.html .
        - signal: bmark-done

In theory (haven't tried it) invoking set-signal inside for-each should work as well.