Canceled timer look like in-progress in durable task dashboard

[wkoeter-bidn]

When using this pattern for timeouts, a canceled timer looks like it is still in progress.

https://learn.microsoft.com/en-us/azure/azure-functions/durable/durable-functions-timers?tabs=csharp#usage-for-timeout

var o1task = context.CallSubOrchestratorAsync<o1Response>("o1",
    new o1Request();

using (var notificationCts = new CancellationTokenSource())
{
    var timeoutTask = context.CreateTimer(context.CurrentUtcDateTime.AddMinutes(2), notificationCts.Token);

    var completedTask = await Task.WhenAny(o1task, timeoutTask);
    if (completedTask == o1task)
    {
        notificationCts.Cancel();
    }
    else
    {
        var p = await context.CallActivityAsync<string>(nameof(NotifyLongRunningActivity), "");
    }
}

var o1 = await o1task;

After cancelling the timer, the task remains with the "in progress" symbol and the timeline remains active.

[philliphoff]

@wkoeter-bidn As the docs explain, an orchestration implementation must "cancel" any outstanding timers before the orchestration can transition to the completed state. This is so that the worker knows the timer is no longer needed and will not be awaited at some later point. However, from the scheduler perspective, timers are actually never canceled (except implicitly, should the timer reaches its terminal state). That is, cancelation is purely a construct of the worker.

What you see in the dashboard timeline is that discrepency between the scheduler and the worker. The scheduler has no knowledge of the orchestration implementation or whether a timer is still "in use" (or not). The timer exists until it either fires or the orchestration completes (or fails).

Maybe there is argument for the worker informing the scheduler of timer cancellations but, off the top of my head, I don't know what implications that might have.

[wkoeter-bidn]

@philliphoff thank you for the explanation.

The dashboard is neat, and we love it. However, for anybody inspecting the current state of a flow, it can be misleading to see a line still being active, while it is in fact inactive. A timer potentially still being active and possibly triggering might have real implications.

[philliphoff]

@wkoeter-bidn No, I get it. I was just pointing out that the timeline is correct from the perspective of DTS, given that it has no knowledge of how a timer is used by the worker's orchestration. I've posted a question in the Durable Task .NET SDK repo (durabletask-dotnet#480) to see if anyone has some insight into this.