Monitor pattern

The monitor pattern is a recurring process in a workflow that polls an external system until a condition is met. For example, it checks job status until it completes, or watches weather data until skies are clear. Unlike a fixed-schedule timer trigger, a monitor waits between iterations (avoiding overlap), supports dynamic intervals, and can terminate itself once the condition is satisfied or a timeout expires.

This article explains how to implement the monitor pattern by using durable orchestrations.

Tip

This article shows the complete implementation. For a conceptual overview of durable orchestration use cases, see What is Durable Task?

The Durable Functions examples include a weather monitoring scenario (C#/JavaScript) and a GitHub issue monitoring scenario (Python).

Note

Version 4 of the Node.js programming model for Azure Functions is generally available. The v4 model is designed to provide a more flexible and intuitive experience for JavaScript and TypeScript developers. For more information about the differences between v3 and v4, see the migration guide.

In the following code snippets, JavaScript (PM4) denotes programming model v4, the new experience.

The Durable Task SDKs example demonstrates job status monitoring with configurable polling intervals by using .NET, JavaScript, Python, and Java.

Prerequisites

.NET 8.0 SDK or later
Access to Azure Durable Task Scheduler or the local emulator

Monitor scenario overview

This sample monitors a location's current weather conditions and alerts a user by SMS when the skies are clear. You could use a regular timer-triggered function to check the weather and send alerts. However, one problem with this approach is lifetime management. If only one alert should be sent, the monitor needs to disable itself after clear weather is detected.

The monitoring pattern can end its own execution, among other benefits:

Monitors run on intervals, not schedules: a timer trigger runs every hour; a monitor waits one hour between actions. A monitor's actions don't overlap unless you specify otherwise, which can be important for long-running tasks.
Monitors can have dynamic intervals: the wait time can change based on some condition.
Monitors can terminate when some condition is met or be terminated by another process.
Monitors can take parameters. The sample shows how the same monitoring process can be applied to any requested location, phone number, or repository.
Monitors are scalable. Because each monitor is an orchestration instance, you can create multiple monitors without having to create new functions or define more code.
Monitors integrate easily into larger workflows. A monitor can be one section of a more complex orchestration function, or a sub-orchestration.

This sample monitors the status of a long-running job and returns the final result when the job completes or times out. You could use a regular polling loop to check job status, but this approach has limitations around lifetime management and reliability.

The monitoring pattern provides these benefits:

Durable polling: The orchestration survives process restarts and can continue monitoring even after failures.
Configurable intervals: You can adjust the wait time between status checks dynamically.
Timeout support: The monitor can terminate when a condition is met or a timeout expires.
Status visibility: Clients can query the orchestration's custom status to see current monitoring progress.
Scalability: Multiple monitors can run concurrently, each tracking different jobs.

Configuration

Configuring a weather API

The C#/JavaScript samples call a weather API to check current conditions. You need to provide your own weather API key and update the sample code accordingly. The sample code references a WeatherUndergroundApiKey app setting - replace this key with your chosen weather provider's key.

App setting name	Value description
WeatherUndergroundApiKey	Your weather API key (replace with your provider's key name as needed).

Configuring a weather API

App setting name	Value description
WeatherUndergroundApiKey	Your weather API key (replace with your provider's key name as needed).

Orchestrator

using Microsoft.Azure.Functions.Worker;
using Microsoft.DurableTask;
using Microsoft.Extensions.Logging;

namespace VSSample;

public static partial class Monitor
{
    [Function("E3_Monitor")]
    public static async Task Run(
        [OrchestrationTrigger] TaskOrchestrationContext context)
    {
        MonitorRequest input = context.GetInput<MonitorRequest>()
            ?? throw new ArgumentNullException(nameof(context), "An input object is required.");
        VerifyRequest(input);

        ILogger logger = context.CreateReplaySafeLogger("E3_Monitor");
        DateTime endTime = context.CurrentUtcDateTime.AddHours(6);
        logger.LogInformation(
            "Instantiating monitor for {Location}. Expires: {EndTime}.",
            input.Location,
            endTime);

        while (context.CurrentUtcDateTime < endTime)
        {
            logger.LogInformation(
                "Checking current weather conditions for {Location} at {CurrentTime}.",
                input.Location,
                context.CurrentUtcDateTime);

            bool isClear = await context.CallActivityAsync<bool>(
                "E3_GetIsClear",
                input.Location);

            if (isClear)
            {
                await context.CallActivityAsync(
                    "E3_SendGoodWeatherAlert",
                    input.Phone);
                break;
            }

            DateTime nextCheckpoint = context.CurrentUtcDateTime.AddMinutes(30);
            await context.CreateTimer(nextCheckpoint, CancellationToken.None);
        }

        logger.LogInformation("Monitor expiring.");
    }

    private static void VerifyRequest(MonitorRequest request)
    {
        ArgumentNullException.ThrowIfNull(request.Location);
        ArgumentException.ThrowIfNullOrEmpty(request.Phone);
    }
}

public sealed class MonitorRequest
{
    public required Location Location { get; init; }

    public required string Phone { get; init; }
}

public sealed class Location
{
    public required string State { get; init; }

    public required string City { get; init; }

    public override string ToString() => $"{City}, {State}";
}

[FunctionName("E3_Monitor")]
public static async Task Run([OrchestrationTrigger] IDurableOrchestrationContext monitorContext, ILogger log)
{
    MonitorRequest input = monitorContext.GetInput<MonitorRequest>();
    if (!monitorContext.IsReplaying) { log.LogInformation($"Received monitor request. Location: {input?.Location}. Phone: {input?.Phone}."); }

    VerifyRequest(input);

    DateTime endTime = monitorContext.CurrentUtcDateTime.AddHours(6);
    if (!monitorContext.IsReplaying) { log.LogInformation($"Instantiating monitor for {input.Location}. Expires: {endTime}."); }

    while (monitorContext.CurrentUtcDateTime < endTime)
    {
        // Check the weather
        if (!monitorContext.IsReplaying) { log.LogInformation($"Checking current weather conditions for {input.Location} at {monitorContext.CurrentUtcDateTime}."); }

        bool isClear = await monitorContext.CallActivityAsync<bool>("E3_GetIsClear", input.Location);

        if (isClear)
        {
            // It's not raining! Or snowing. Or misting. Tell our user to take advantage of it.
            if (!monitorContext.IsReplaying) { log.LogInformation($"Detected clear weather for {input.Location}. Notifying {input.Phone}."); }

            await monitorContext.CallActivityAsync("E3_SendGoodWeatherAlert", input.Phone);
            break;
        }
        else
        {
            // Wait for the next checkpoint
            var nextCheckpoint = monitorContext.CurrentUtcDateTime.AddMinutes(30);
            if (!monitorContext.IsReplaying) { log.LogInformation($"Next check for {input.Location} at {nextCheckpoint}."); }

            await monitorContext.CreateTimer(nextCheckpoint, CancellationToken.None);
        }
    }

    log.LogInformation($"Monitor expiring.");
}

[Deterministic]
private static void VerifyRequest(MonitorRequest request)
{
    if (request == null)
    {
        throw new ArgumentNullException(nameof(request), "An input object is required.");
    }

    if (request.Location == null)
    {
        throw new ArgumentNullException(nameof(request.Location), "A location input is required.");
    }

    if (string.IsNullOrEmpty(request.Phone))
    {
        throw new ArgumentNullException(nameof(request.Phone), "A phone number input is required.");
    }
}

The orchestrator function requires a location to monitor and a phone number to send a message to when the weather becomes clear at the location. You pass this data to the orchestrator function as a strongly typed MonitorRequest object.

The E3_Monitor function uses the standard function.json for orchestrator functions.

{
  "bindings": [
    {
      "name": "context",
      "type": "orchestrationTrigger",
      "direction": "in"
    }
  ],
  "disabled": false
}

Here's the code that implements the function:

const df = require("durable-functions");
const moment = require("moment");

module.exports = df.orchestrator(function* (context) {
    const input = context.df.getInput();
    context.log(
        "Received monitor request. location: " +
            (input ? input.location : undefined) +
            ". phone: " +
            (input ? input.phone : undefined) +
            "."
    );

    verifyRequest(input);

    const endTime = moment.utc(context.df.currentUtcDateTime).add(6, "h");
    context.log(
        "Instantiating monitor for " +
            input.location.city +
            ", " +
            input.location.state +
            ". Expires: " +
            endTime +
            "."
    );

    while (moment.utc(context.df.currentUtcDateTime).isBefore(endTime)) {
        // Check the weather
        context.log(
            "Checking current weather conditions for " +
                input.location.city +
                ", " +
                input.location.state +
                " at " +
                context.df.currentUtcDateTime +
                "."
        );
        const isClear = yield context.df.callActivity("E3_GetIsClear", input.location);

        if (isClear) {
            // It's not raining! Or snowing. Or misting. Tell our user to take advantage of it.
            context.log(
                "Detected clear weather for " +
                    input.location.city +
                    ", " +
                    input.location.state +
                    ". Notifying " +
                    input.phone +
                    "."
            );

            yield context.df.callActivity("E3_SendGoodWeatherAlert", input.phone);
            break;
        } else {
            // Wait for the next checkpoint
            const nextCheckpoint = moment.utc(context.df.currentUtcDateTime).add(30, "s");
            context.log(
                "Next check for " +
                    input.location.city +
                    ", " +
                    input.location.state +
                    " at " +
                    nextCheckpoint.toString()
            );

            yield context.df.createTimer(nextCheckpoint.toDate()); // accomodate cancellation tokens
        }
    }

    context.log("Monitor expiring.");
});

function verifyRequest(request) {
    if (!request) {
        throw new Error("An input object is required.");
    }
    if (!request.location) {
        throw new Error("A location input is required.");
    }
    if (!request.phone) {
        throw new Error("A phone number input is required.");
    }
}

The E3_Monitor function uses the standard function.json for orchestrator functions.

{
  "scriptFile": "__init__.py",
  "bindings": [
    {
      "name": "context",
      "type": "orchestrationTrigger",
      "direction": "in"
    }
  ]
}

Here's the code that implements the function:

import azure.durable_functions as df
from datetime import timedelta
from typing import Dict

def orchestrator_function(context: df.DurableOrchestrationContext):

    monitoring_request: Dict[str, str] = context.get_input()
    repo_url: str = monitoring_request["repo"]
    phone: str = monitoring_request["phone"]

    # Expiration of the repo monitoring
    expiry_time = context.current_utc_datetime + timedelta(minutes=5)
    while context.current_utc_datetime < expiry_time:
        # Count the number of issues in the repo (the GitHub API caps at 30 issues per page)
        too_many_issues = yield context.call_activity("E3_TooManyOpenIssues", repo_url)

        # If we detect too many issues, we text the provided phone number
        if too_many_issues:
            # Extract URLs of GitHub issues, and return them
            yield context.call_activity("E3_SendAlert", phone)
            break
        else:

            # Reporting the number of statuses found
            status = f"The repository does not have too many issues, for now ..."
            context.set_custom_status(status)

            # Schedule a new "wake up" signal
            next_check = context.current_utc_datetime + timedelta(minutes=1)
            yield context.create_timer(next_check)

    return "Monitor completed!"

main = df.Orchestrator.create(orchestrator_function)

The orchestrator function periodically checks a condition and sends an alert if it's met. It uses a durable timer for polling intervals and continues until the monitor expires.

param($Context)

$input = $Context.Input | ConvertFrom-Json
$expirationTime = (Get-Date).AddHours(6)
$pollingInterval = New-TimeSpan -Seconds $input.pollingIntervalSeconds

while ((Get-Date) -lt $expirationTime) {
    # Check current conditions
    $isClear = Invoke-DurableActivity -FunctionName 'E3_GetIsClear' -Input $input.location

    if ($isClear) {
        # Condition met - send alert and exit
        Invoke-DurableActivity -FunctionName 'E3_SendGoodWeatherAlert' -Input $input.phone
        break
    }

    # Wait for the next polling interval
    Start-DurableTimer -Duration $pollingInterval
}

@FunctionName("E3_Monitor")
public void monitorOrchestrator(
        @DurableOrchestrationTrigger(name = "ctx") TaskOrchestrationContext ctx) {
    MonitorRequest input = ctx.getInput(MonitorRequest.class);
    Instant expirationTime = ctx.getCurrentInstant().plus(Duration.ofHours(6));
    int pollingInterval = input.getPollingIntervalSeconds();

    while (ctx.getCurrentInstant().isBefore(expirationTime)) {
        // Check current conditions
        boolean isClear = ctx.callActivity(
            "E3_GetIsClear", input.getLocation(), boolean.class).await();

        if (isClear) {
            // Condition met - send alert and exit
            ctx.callActivity("E3_SendGoodWeatherAlert", input.getPhone()).await();
            break;
        }

        // Wait for the next polling interval
        Instant nextCheck = ctx.getCurrentInstant().plus(
            Duration.ofSeconds(pollingInterval));
        ctx.createTimer(nextCheck).await();
    }
}

This orchestrator function performs the following actions:

Gets the MonitorRequest consisting of the location to monitor and the phone number to which it sends an SMS notification (or repo for the Python example).
Determines the expiration time of the monitor. The sample uses a hard-coded value for brevity.
Calls the status-checking activity to determine whether the condition is met.
If the condition is met, calls the alert activity to send a notification.
Creates a durable timer to resume the orchestration at the next polling interval. The sample uses a hard-coded value for brevity.
Continues running until the current UTC time passes the monitor's expiration time, or an alert is sent.

You can run multiple orchestrator function instances simultaneously by calling the orchestrator function multiple times. You can specify the location to monitor and the phone number to send an alert to. The orchestrator function isn't running while waiting for the timer, so you aren't charged for it.

The orchestrator periodically checks the status of a job and returns when the job completes or times out.

using Microsoft.DurableTask;
using System;
using System.Threading.Tasks;

[DurableTask(nameof(MonitoringJobOrchestration))]
public class MonitoringJobOrchestration : TaskOrchestrator<JobMonitorInput, JobMonitorResult>
{
    public override async Task<JobMonitorResult> RunAsync(
        TaskOrchestrationContext context, JobMonitorInput input)
    {
        var jobId = input.JobId;
        var pollingInterval = TimeSpan.FromSeconds(input.PollingIntervalSeconds);
        var expirationTime = context.CurrentUtcDateTime.AddSeconds(input.TimeoutSeconds);

        // Initialize monitoring state
        int checkCount = 0;

        while (context.CurrentUtcDateTime < expirationTime)
        {
            // Check current job status
            var jobStatus = await context.CallActivityAsync<JobStatus>(
                nameof(CheckJobStatusActivity),
                new CheckJobInput { JobId = jobId, CheckCount = checkCount });

            checkCount = jobStatus.CheckCount;

            // Make job status available via custom status
            context.SetCustomStatus(jobStatus);

            if (jobStatus.Status == "Completed")
            {
                return new JobMonitorResult
                {
                    JobId = jobId,
                    FinalStatus = "Completed",
                    ChecksPerformed = checkCount
                };
            }

            // Calculate next check time
            var nextCheck = context.CurrentUtcDateTime.Add(pollingInterval);
            if (nextCheck > expirationTime)
            {
                nextCheck = expirationTime;
            }

            // Wait until next polling interval
            await context.CreateTimer(nextCheck, default);
        }

        // Timeout reached
        return new JobMonitorResult
        {
            JobId = jobId,
            FinalStatus = "Timeout",
            ChecksPerformed = checkCount
        };
    }
}

import {
  OrchestrationContext,
  TOrchestrator,
} from "@microsoft/durabletask-js";

const monitorOrchestrator: TOrchestrator = async function* (
  ctx: OrchestrationContext,
  input: { jobId: string; pollingIntervalSeconds: number; timeoutSeconds: number }
): any {
  const { jobId, pollingIntervalSeconds, timeoutSeconds } = input;
  const expirationTime = new Date(
    ctx.currentUtcDateTime.getTime() + timeoutSeconds * 1000
  );

  let checkCount = 0;

  while (ctx.currentUtcDateTime < expirationTime) {
    // Check current job status
    const jobStatus: any = yield ctx.callActivity(checkJobStatus, {
      jobId,
      checkCount,
    });

    checkCount = jobStatus.checkCount;

    // Make job status available via custom status
    ctx.setCustomStatus(jobStatus);

    if (jobStatus.status === "Completed") {
      return {
        jobId,
        finalStatus: "Completed",
        checksPerformed: checkCount,
      };
    }

    // Wait for next polling interval
    yield ctx.createTimer(pollingIntervalSeconds);
  }

  // Timeout reached
  return {
    jobId,
    finalStatus: "Timeout",
    checksPerformed: checkCount,
  };
};

import datetime
from durabletask import task

def monitoring_job_orchestrator(ctx: task.OrchestrationContext, job_data: dict) -> dict:
    """
    Orchestrator that demonstrates the monitoring pattern.

    Periodically checks the status of a job until it completes or times out.
    """
    job_id = job_data.get("job_id")
    polling_interval = job_data.get("polling_interval_seconds", 5)
    timeout = job_data.get("timeout_seconds", 30)

    # Record the start time
    start_time = ctx.current_utc_datetime
    expiration_time = start_time + datetime.timedelta(seconds=timeout)

    # Initialize monitoring state
    job_status = {
        "job_id": job_id,
        "status": "Unknown",
        "check_count": 0
    }

    # Loop until the job completes or times out
    while True:
        # Check current job status
        check_input = {"job_id": job_id, "check_count": job_status.get("check_count", 0)}
        job_status = yield ctx.call_activity("check_job_status", input=check_input)

        # Make the job status available via custom status
        ctx.set_custom_status(job_status)

        if job_status["status"] == "Completed":
            break

        # Check if we've hit the timeout
        current_time = ctx.current_utc_datetime
        if current_time >= expiration_time:
            job_status["status"] = "Timeout"
            break

        # Calculate next check time
        next_check_time = current_time + datetime.timedelta(seconds=polling_interval)
        if next_check_time > expiration_time:
            next_check_time = expiration_time

        # Wait until next polling interval
        yield ctx.create_timer(next_check_time)

    # Return the final status
    return {
        "job_id": job_id,
        "final_status": job_status["status"],
        "checks_performed": job_status["check_count"]
    }

import com.microsoft.durabletask.*;
import com.microsoft.durabletask.azuremanaged.DurableTaskSchedulerWorkerExtensions;
import java.time.Duration;

DurableTaskGrpcWorker worker = DurableTaskSchedulerWorkerExtensions.createWorkerBuilder(connectionString)
    .addOrchestration(new TaskOrchestrationFactory() {
        @Override
        public String getName() { return "MonitoringJobOrchestrator"; }

        @Override
        public TaskOrchestration create() {
            return ctx -> {
                JobData jobData = ctx.getInput(JobData.class);
                int pollingCount = 0;

                // Set initial status
                ctx.setCustomStatus(new JobStatus("Starting monitoring..."));

                while (true) {
                    // Update status
                    ctx.setCustomStatus(new JobStatus(
                        "Polling job status (attempt " + (++pollingCount) + ")"));

                    // Wait for polling interval
                    ctx.createTimer(Duration.ofSeconds(jobData.pollingIntervalSeconds)).await();

                    // Check if job is complete (simulated after 3 attempts)
                    if (pollingCount >= 3) {
                        ctx.setCustomStatus(new JobStatus("Job completed successfully"));
                        ctx.complete(new JobResult(
                            "COMPLETED",
                            "Job completed after " + pollingCount + " attempts"));
                        break;
                    }
                }
            };
        }
    })
    .build();

This orchestrator performs the following actions:

Takes the job ID, polling interval, and timeout as input parameters.
Records the start time and calculates the expiration time.
Enters a polling loop that checks the job status.
Updates the custom status so clients can monitor progress.
If the job completes, returns the final result.
If the timeout is reached, returns a timeout status.
Uses CreateTimer to wait between polling attempts without consuming resources.

Activities

As with other samples, the helper activity functions are regular functions that use the activityTrigger trigger binding.

Status checking activity

The E3_GetIsClear function gets the current weather conditions by using the Weather Underground API and determines whether the sky is clear.

        [FunctionName("E3_GetIsClear")]
        public static async Task<bool> GetIsClear([ActivityTrigger] Location location)
        {
            var currentConditions = await WeatherUnderground.GetCurrentConditionsAsync(location);
            return currentConditions.Equals(WeatherCondition.Clear);
        }

The function.json is defined as follows:

{
  "bindings": [
    {
      "name": "location",
      "type": "activityTrigger",
      "direction": "in"
    }
  ],
  "disabled": false
}

Here's the implementation.

const request = require("request");

const clearWeatherConditions = [
    "Overcast",
    "Clear",
    "Partly Cloudy",
    "Mostly Cloudy",
    "Scattered Clouds",
];

module.exports = function (context, location) {
    getCurrentConditions(location)
        .then(function (data) {
            const isClear = clearWeatherConditions.includes(data.weather);
            context.done(null, isClear);
        })
        .catch(function (err) {
            context.log(`E3_GetIsClear encountered an error: ${err}`);
            context.done(err);
        });
};

function getCurrentConditions(location) {
    return new Promise(function (resolve, reject) {
        const options = {
            url: `https://api.wunderground.com/api/${process.env["WeatherUndergroundApiKey"]}/conditions/q/${location.state}/${location.city}.json`,
            method: "GET",
            json: true,
        };
        request(options, function (err, res, body) {
            if (err) {
                reject(err);
            }
            if (body.error) {
                reject(body.error);
            }
            if (body.response.error) {
                reject(body.response.error);
            }
            resolve(body.current_observation);
        });
    });
}

The E3_TooManyOpenIssues function gets a list of currently open issues on the repo and determines if there are "too many" of them: more than 3 as per the sample.

The function.json is defined as follows:

{
  "scriptFile": "__init__.py",
  "bindings": [
    {
      "name": "repoID",
      "type": "activityTrigger",
      "direction": "in"
    }
  ]
}

Here's the implementation.

import requests
import json

def main(repoID: str) -> str:

    # We use the GitHub API to count the number of open issues in the repo provided
    # Note that the GitHub API only displays at most 30 issues per response, so
    # the maximum number this activity will return is 30. That's enough for demo'ing purposes.
    [user, repo] = repoID.split("/")
    url = f"https://api.github.com/repos/{user}/{repo}/issues?state=open"
    res = requests.get(url)
    if res.status_code != 200:
        error_message = f"Could not find repo {user} under {repo}! API endpoint hit was: {url}"
        raise Exception(error_message)
    issues = json.loads(res.text)
    too_many_issues: bool = len(issues) >= 3
    return too_many_issues

The E3_GetIsClear function checks whether the condition is met. In this example, it checks weather conditions for a location.

param($location)

# Call weather API to check current conditions
# In a real app, call an external API here
$conditions = Get-WeatherConditions -Location $location
$conditions -eq 'Clear'

The E3_GetIsClear function checks the current weather conditions for a location by using a weather API.

@FunctionName("E3_GetIsClear")
public boolean getIsClear(
        @DurableActivityTrigger(name = "location") Location location) {
    // Call weather API to check current conditions
    String conditions = getWeatherConditions(location);
    return conditions.equals("Clear");
}

Run the monitor sample

By using the HTTP-triggered functions included in the sample, you can start the orchestration by sending the following HTTP POST request:

POST https://{host}/orchestrators/E3_Monitor
Content-Length: 77
Content-Type: application/json

{ "location": { "city": "Redmond", "state": "WA" }, "phone": "+1425XXXXXXX" }

By using the HTTP-triggered functions included in the sample, you can start the orchestration by sending the following HTTP POST request:

POST https://{host}/orchestrators/E3_Monitor
Content-Length: 77
Content-Type: application/json

{ "location": { "city": "Redmond", "state": "WA" }, "phone": "+1425XXXXXXX" }

You need a GitHub account. Create a temporary public repository that you can open issues to.

By using the HTTP-triggered functions included in the sample, you can start the orchestration by sending the following HTTP POST request:

POST https://{host}/orchestrators/E3_Monitor
Content-Length: 77
Content-Type: application/json

{ "repo": "<your GitHub handle>/<a new GitHub repo under your user>", "phone": "+1425XXXXXXX" }

For example, if your GitHub username is foo and your repository is bar, set the value for "repo" to "foo/bar".

By using the HTTP-triggered function included in the sample, you can start the orchestration by sending the following HTTP POST request:

POST https://{host}/api/orchestrators/E3_Monitor
Content-Type: application/json

{ "location": { "city": "Redmond", "state": "WA" }, "phone": "+1425XXXXXXX" }

By using the HTTP-triggered function included in the sample, you can start the orchestration by sending the following HTTP POST request:

POST https://{host}/api/StartMonitor
Content-Type: application/json

{ "location": { "city": "Redmond", "state": "WA" }, "phone": "+1425XXXXXXX" }

The HTTP trigger function schedules the orchestration:

@FunctionName("StartMonitor")
public HttpResponseMessage startMonitor(
        @HttpTrigger(name = "req", methods = {HttpMethod.POST}) HttpRequestMessage<Optional<String>> req,
        @DurableClientInput(name = "durableContext") DurableClientContext durableContext,
        final ExecutionContext context) {
    DurableTaskClient client = durableContext.getClient();
    String instanceId = client.scheduleNewOrchestrationInstance("E3_Monitor", req.getBody().get());
    context.getLogger().info("Started monitor orchestration with ID = " + instanceId);
    return durableContext.createCheckStatusResponse(req, instanceId);
}

HTTP/1.1 202 Accepted
Content-Type: application/json; charset=utf-8
Location: https://{host}/runtime/webhooks/durabletask/instances/f6893f25acf64df2ab53a35c09d52635?taskHub=SampleHubVS&connection=Storage&code={SystemKey}
RetryAfter: 10

{"id": "f6893f25acf64df2ab53a35c09d52635", "statusQueryGetUri": "https://{host}/runtime/webhooks/durabletask/instances/f6893f25acf64df2ab53a35c09d52635?taskHub=SampleHubVS&connection=Storage&code={systemKey}", "sendEventPostUri": "https://{host}/runtime/webhooks/durabletask/instances/f6893f25acf64df2ab53a35c09d52635/raiseEvent/{eventName}?taskHub=SampleHubVS&connection=Storage&code={systemKey}", "terminatePostUri": "https://{host}/runtime/webhooks/durabletask/instances/f6893f25acf64df2ab53a35c09d52635/terminate?reason={text}&taskHub=SampleHubVS&connection=Storage&code={systemKey}"}

The E3_Monitor instance starts and queries the current conditions. If the condition is met, it calls an activity function to send an alert; otherwise, it sets a timer. When the timer expires, the orchestration resumes.

You can see the orchestration's activity by looking at the function logs in the Azure Functions portal.

The orchestration completes once its timeout is reached or the condition is detected. You can also use the terminate API inside another function or invoke the terminatePostUri HTTP POST webhook referenced in the preceding 202 response. To use the webhook, replace {text} with the reason for the early termination. The HTTP POST URL looks roughly as follows:

POST https://{host}/runtime/webhooks/durabletask/instances/f6893f25acf64df2ab53a35c09d52635/terminate?reason=Because&taskHub=SampleHubVS&connection=Storage&code={systemKey}

To run the sample, you need:

Start the Durable Task Scheduler emulator (for local development):

docker run -d -p 8080:8080 -p 8082:8082 --name dts-emulator mcr.microsoft.com/dts/dts-emulator:latest

Start the worker to register the orchestrator and activities.
Run the client to schedule a monitoring orchestration.

using System;
using System.Threading.Tasks;

var client = DurableTaskClientBuilder.UseDurableTaskScheduler(connectionString).Build();

// Schedule the monitoring orchestration
var input = new JobMonitorInput
{
    JobId = "job-" + Guid.NewGuid().ToString(),
    PollingIntervalSeconds = 5,
    TimeoutSeconds = 30
};

string instanceId = await client.ScheduleNewOrchestrationInstanceAsync(
    nameof(MonitoringJobOrchestration), input);

Console.WriteLine($"Started monitoring orchestration: {instanceId}");

// Wait for completion while checking status
while (true)
{
    var state = await client.GetInstanceMetadataAsync(instanceId, getInputsAndOutputs: true);

    if (state.RuntimeStatus == OrchestrationRuntimeStatus.Completed ||
        state.RuntimeStatus == OrchestrationRuntimeStatus.Failed)
    {
        Console.WriteLine($"Monitoring completed: {state.ReadOutputAs<JobMonitorResult>().FinalStatus}");
        break;
    }

    Console.WriteLine($"Current status: {state.ReadCustomStatusAs<JobStatus>()?.Status}");
    await Task.Delay(2000);
}

import {
  DurableTaskAzureManagedClientBuilder,
  DurableTaskAzureManagedWorkerBuilder,
} from "@microsoft/durabletask-js-azuremanaged";

const client = new DurableTaskAzureManagedClientBuilder()
  .connectionString(connectionString)
  .build();

const worker = new DurableTaskAzureManagedWorkerBuilder()
  .connectionString(connectionString)
  .addOrchestrator(monitorOrchestrator)
  .addActivity(checkJobStatus)
  .build();

await worker.start();

// Schedule the monitoring orchestration
const input = {
  jobId: `job-${Date.now()}`,
  pollingIntervalSeconds: 5,
  timeoutSeconds: 30,
};

const instanceId = await client.scheduleNewOrchestration(
  monitorOrchestrator,
  input
);

console.log(`Started monitoring orchestration: ${instanceId}`);

// Wait for completion
const result = await client.waitForOrchestrationCompletion(
  instanceId,
  true,
  60
);
console.log(`Final result: ${result?.serializedOutput}`);

await worker.stop();
await client.stop();

from durabletask.azuremanaged.client import DurableTaskSchedulerClient
import time

client = DurableTaskSchedulerClient(
    host_address=endpoint,
    secure_channel=endpoint != "http://localhost:8080",
    taskhub=taskhub,
    token_credential=credential
)

# Schedule the monitoring orchestration
job_data = {
    "job_id": "job-123",
    "polling_interval_seconds": 5,
    "timeout_seconds": 30
}

instance_id = client.schedule_new_orchestration(
    monitoring_job_orchestrator,
    input=job_data
)

print(f"Started monitoring orchestration: {instance_id}")

# Wait for completion
result = client.wait_for_orchestration_completion(instance_id, timeout=60)
print(f"Final result: {result.serialized_output}")

import java.time.Duration;
import java.util.UUID;

DurableTaskClient client = DurableTaskSchedulerClientExtensions
    .createClientBuilder(connectionString).build();

// Schedule the monitoring orchestration
JobData jobData = new JobData(
    "job-" + UUID.randomUUID().toString(),
    5,  // polling interval seconds
    30  // timeout seconds
);

String instanceId = client.scheduleNewOrchestrationInstance(
    "MonitoringJobOrchestrator",
    new NewOrchestrationInstanceOptions().setInput(jobData));

System.out.println("Started monitoring orchestration: " + instanceId);

// Wait for completion
OrchestrationMetadata result = client.waitForInstanceCompletion(
    instanceId, Duration.ofSeconds(60), true);

System.out.println("Final result: " + result.readOutputAs(JobResult.class).status);

Next steps

This sample demonstrates how to use Durable Functions to monitor an external source's status by using durable timers and conditional logic. The next sample shows how to use external events and durable timers to handle human interaction.

Run the human interaction sample

This sample demonstrated how to use the Durable Task SDKs to implement the monitoring pattern with durable timers and status tracking. To learn more about other patterns and features, see:

Durable Task JavaScript SDK on GitHub

Get started with Durable Task SDKs

Last updated on 2026-08-14

Monitor pattern

Prerequisites

Monitor scenario overview

Configuration

Configuring a weather API

Orchestrator

Activities

Status checking activity

Run the monitor sample

Next steps

Additional resources