Monitor scenario in Durable Functions - Weather watcher sample

Article
11/24/2022

The monitor pattern refers to a flexible recurring process in a workflow - for example, polling until certain conditions are met. This article explains a sample that uses Durable Functions to implement monitoring.

Prerequisites

C#
JavaScript

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 will not overlap unless specified, 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 weather-monitoring process can be applied to any requested location and phone number.
Monitors are scalable. Because each monitor is an orchestration instance, multiple monitors can be created 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.

Configuration

Configuring Twilio integration

This sample involves using the Twilio service to send SMS messages to a mobile phone. Azure Functions already has support for Twilio via the Twilio binding, and the sample uses that feature.

The first thing you need is a Twilio account. You can create one free at https://www.twilio.com/try-twilio. Once you have an account, add the following three app settings to your function app.

App setting name	Value description
TwilioAccountSid	The SID for your Twilio account
TwilioAuthToken	The Auth token for your Twilio account
TwilioPhoneNumber	The phone number associated with your Twilio account. This is used to send SMS messages.

Configuring Weather Underground integration

This sample involves using the Weather Underground API to check current weather conditions for a location.

The first thing you need is a Weather Underground account. You can create one for free at https://www.wunderground.com/signup. Once you have an account, you will need to acquire an API key. You can do so by visiting https://www.wunderground.com/weather/api, then selecting Key Settings. The Stratus Developer plan is free and sufficient to run this sample.

Once you have an API key, add the following app setting to your function app.

App setting name	Value description
WeatherUndergroundApiKey	Your Weather Underground API key.

The functions

This article explains the following functions in the sample app:

E3_Monitor: An orchestrator function that calls E3_GetIsClear periodically. It calls E3_SendGoodWeatherAlert if E3_GetIsClear returns true.
E3_GetIsClear: An activity function that checks the current weather conditions for a location.
E3_SendGoodWeatherAlert: An activity function that sends an SMS message via Twilio.

[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 requires a location to monitor and a phone number to send a message to when the whether becomes clear at the location. This data is passed to the orchestrator 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 is 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.");
    }
}

This orchestrator function performs the following actions:

Gets the MonitorRequest consisting of the location to monitor and the phone number to which it will send an SMS notification.
Determines the expiration time of the monitor. The sample uses a hard-coded value for brevity.
Calls E3_GetIsClear to determine whether there are clear skies at the requested location.
If the weather is clear, calls E3_SendGoodWeatherAlert to send an SMS notification to the requested phone number.
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 SMS alert is sent.

Multiple orchestrator instances can run simultaneously by calling the orchestrator function multiple times. The location to monitor and the phone number to send an SMS alert to can be specified. Finally, do note that the orchestrator function is not running while waiting for the timer, so you will not get charged for it.

E3_GetIsClear activity function

As with other samples, the helper activity functions are regular functions that use the activityTrigger trigger binding. The E3_GetIsClear function gets the current weather conditions using the Weather Underground API and determines whether the sky is clear.

C#
JavaScript

[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
}

And here is 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);
        });
    });
}

E3_SendGoodWeatherAlert activity function

The E3_SendGoodWeatherAlert function uses the Twilio binding to send an SMS message notifying the end user that it's a good time for a walk.

C#
JavaScript

    [FunctionName("E3_SendGoodWeatherAlert")]
    public static void SendGoodWeatherAlert(
        [ActivityTrigger] string phoneNumber,
        ILogger log,
        [TwilioSms(AccountSidSetting = "TwilioAccountSid", AuthTokenSetting = "TwilioAuthToken", From = "%TwilioPhoneNumber%")]
            out CreateMessageOptions message)
    {
        message = new CreateMessageOptions(new PhoneNumber(phoneNumber));
        message.Body = $"The weather's clear outside! Go take a walk!";
    }

internal class WeatherUnderground
{
    private static readonly HttpClient httpClient = new HttpClient();
    private static IReadOnlyDictionary<string, WeatherCondition> weatherMapping = new Dictionary<string, WeatherCondition>()
    {
        { "Clear", WeatherCondition.Clear },
        { "Overcast", WeatherCondition.Clear },
        { "Cloudy", WeatherCondition.Clear },
        { "Clouds", WeatherCondition.Clear },
        { "Drizzle", WeatherCondition.Precipitation },
        { "Hail", WeatherCondition.Precipitation },
        { "Ice", WeatherCondition.Precipitation },
        { "Mist", WeatherCondition.Precipitation },
        { "Precipitation", WeatherCondition.Precipitation },
        { "Rain", WeatherCondition.Precipitation },
        { "Showers", WeatherCondition.Precipitation },
        { "Snow", WeatherCondition.Precipitation },
        { "Spray", WeatherCondition.Precipitation },
        { "Squall", WeatherCondition.Precipitation },
        { "Thunderstorm", WeatherCondition.Precipitation },
    };

    internal static async Task<WeatherCondition> GetCurrentConditionsAsync(Location location)
    {
        var apiKey = Environment.GetEnvironmentVariable("WeatherUndergroundApiKey");
        if (string.IsNullOrEmpty(apiKey))
        {
            throw new InvalidOperationException("The WeatherUndergroundApiKey environment variable was not set.");
        }

        var callString = string.Format("http://api.wunderground.com/api/{0}/conditions/q/{1}/{2}.json", apiKey, location.State, location.City);
        var response = await httpClient.GetAsync(callString);
        var conditions = await response.Content.ReadAsAsync<JObject>();

        JToken currentObservation;
        if (!conditions.TryGetValue("current_observation", out currentObservation))
        {
            JToken error = conditions.SelectToken("response.error");

            if (error != null)
            {
                throw new InvalidOperationException($"API returned an error: {error}.");
            }
            else
            {
                throw new ArgumentException("Could not find weather for this location. Try being more specific.");
            }
        }

        return MapToWeatherCondition((string)(currentObservation as JObject).GetValue("weather"));
    }

    private static WeatherCondition MapToWeatherCondition(string weather)
    {
        foreach (var pair in weatherMapping)
        {
            if (weather.Contains(pair.Key))
            {
                return pair.Value;
            }
        }

        return WeatherCondition.Other;
    }
}

Note

You will need to install the Microsoft.Azure.WebJobs.Extensions.Twilio Nuget package to run the sample code.

Its function.json is simple:

{
    "bindings": [
      {
        "name": "phoneNumber",
        "type": "activityTrigger",
        "direction": "in"
      },
      {
        "type": "twilioSms",
        "name": "message",
        "from": "%TwilioPhoneNumber%",
        "accountSidSetting": "TwilioAccountSid",
        "authTokenSetting": "TwilioAuthToken",
        "direction": "out"
      }
    ],
    "disabled": false
  }

And here is the code that sends the SMS message:

module.exports = function (context, phoneNumber) {
    context.bindings.message = {
        body: `The weather's clear outside! Go take a walk!`,
        to: phoneNumber,
    };

    context.done();
};

Run the sample

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" }

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 weather conditions for the requested location. If the weather is clear, it calls an activity function to send an alert; otherwise, it sets a timer. When the timer expires, the orchestration will resume.

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

2018-03-01T01:14:41.649 Function started (Id=2d5fcadf-275b-4226-a174-f9f943c90cd1)
2018-03-01T01:14:42.741 Started orchestration with ID = '1608200bb2ce4b7face5fc3b8e674f2e'.
2018-03-01T01:14:42.780 Function completed (Success, Id=2d5fcadf-275b-4226-a174-f9f943c90cd1, Duration=1111ms)
2018-03-01T01:14:52.765 Function started (Id=b1b7eb4a-96d3-4f11-a0ff-893e08dd4cfb)
2018-03-01T01:14:52.890 Received monitor request. Location: Redmond, WA. Phone: +1425XXXXXXX.
2018-03-01T01:14:52.895 Instantiating monitor for Redmond, WA. Expires: 3/1/2018 7:14:52 AM.
2018-03-01T01:14:52.909 Checking current weather conditions for Redmond, WA at 3/1/2018 1:14:52 AM.
2018-03-01T01:14:52.954 Function completed (Success, Id=b1b7eb4a-96d3-4f11-a0ff-893e08dd4cfb, Duration=189ms)
2018-03-01T01:14:53.226 Function started (Id=80a4cb26-c4be-46ba-85c8-ea0c6d07d859)
2018-03-01T01:14:53.808 Function completed (Success, Id=80a4cb26-c4be-46ba-85c8-ea0c6d07d859, Duration=582ms)
2018-03-01T01:14:53.967 Function started (Id=561d0c78-ee6e-46cb-b6db-39ef639c9a2c)
2018-03-01T01:14:53.996 Next check for Redmond, WA at 3/1/2018 1:44:53 AM.
2018-03-01T01:14:54.030 Function completed (Success, Id=561d0c78-ee6e-46cb-b6db-39ef639c9a2c, Duration=62ms)

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

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

Next steps

This sample has demonstrated how to use Durable Functions to monitor an external source's status 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