Skip to content
TheChatArchiveThe Chat Archive
fable-5· unknownHugging Face ibragim-bad/swe_rebench_07_2026_trajectories

The TriggerSchedulerMonitor reports realtime triggers as locked Describe the issue The…

  1. User

    The TriggerSchedulerMonitor reports realtime triggers as locked

    Describe the issue

    The TriggerSchedulerMonitor reports realtime triggers as locked.

    [tenant: main] [namespace: company.team] [flow: realtime-kafka] [trigger: kafka] No execution found, schedule is blocked since '2026-04-27T13:38:23.415498Z'

    But as those trigger will always be blocked without any executionID they will be reported indefinitly.

    Environment

    • Kestra Version: develop

    Interface

    No new interfaces are introduced.

  2. Assistant

    Now a unit test. Let me check the scheduler test setup and DefaultScheduler.currentVNodesAssignment().

  3. Assistant

    Waiting for the test run to finish.

  4. Assistant

    The test run is executing in the background; I'll report as soon as it completes.

  5. Assistant

    The change is complete pending test verification. Summary of what was done:

    Root cause: TriggerSchedulerMonitor iterates over all locked triggers and warns when no execution is found. Realtime triggers are locked for their entire lifetime by design (they run continuously on a worker and produce executions asynchronously per event), so they were reported as "blocked" indefinitely.

    Fix in scheduler/src/main/java/io/kestra/scheduler/TriggerSchedulerMonitor.java:57-62: filter out TriggerType.REALTIME states before monitoring, so only schedule/polling triggers are checked for missing executions.

    Test added at scheduler/src/test/java/io/kestra/scheduler/TriggerSchedulerMonitorTest.java:

    • shouldNotReportRealtimeTriggerAsBlocked — a locked realtime trigger never triggers an execution lookup (no warning, no missing-execution metric).
    • shouldCheckExecutionForNonRealtimeLockedTrigger — a locked schedule trigger is still checked, preserving existing behavior.

    The Gradle test run is still executing in the background — I'll confirm the results when it finishes.

  6. Assistant

    Gradle is still running (first compile of the scheduler module takes a while; output is piped through tail -30 so it stays empty until completion). The background task will notify me when it exits — no need to re-arm the monitor.

  7. Assistant

    Build successful — both tests pass (tests="2" failures="0" errors="0"). The captured stderr even confirms the behavior split: the SCHEDULE trigger still produces the "schedule is blocked" warning, while the REALTIME trigger produces nothing and never touches the execution repository.

    Done. Final state of the fix:

    • scheduler/src/main/java/io/kestra/scheduler/TriggerSchedulerMonitor.java:57-62 — locked triggers with TriggerType.REALTIME are filtered out before monitoring, since realtime triggers are locked for their whole lifetime by design and create executions asynchronously; they were previously reported as blocked indefinitely.
    • scheduler/src/test/java/io/kestra/scheduler/TriggerSchedulerMonitorTest.java — new unit test with two cases:
      • shouldNotReportRealtimeTriggerAsBlocked — verifies no execution lookup (and hence no warning/metric) for a locked realtime trigger.
      • shouldCheckExecutionForNonRealtimeLockedTrigger — verifies schedule triggers are still monitored as before.

Responses were produced in different, uncontrolled contexts (system prompts, settings, prior turns). This is not a benchmark.