Skip to content

Latest commit

 

History

History
79 lines (51 loc) · 5.98 KB

README.md

File metadata and controls

79 lines (51 loc) · 5.98 KB
Raw

Azure Cosmos DB + Azure Functions: Estimating pending lag - Azure Dev .NET Days

Watch the recording! 👇

Watch the presentation

Description

This repository contains the code reference to help create an Azure Dashboard to track the pending work on a deployed CosmosDBTrigger. The theme is a system that processes device telemetry and produces a real time summary as telemetry is stored in an Azure Cosmos DB container.

It contains three projects:

  1. cosmos.trigger: A CosmosDBTrigger with retry policy support, that generates a summary per device and stores it on another Cosmos DB container.
  2. cosmos.monitor: A TimerTrigger that will leverage the Change Feed Estimator, obtain lag estimations across leases, and report it as Metrics on Application Insights.
  3. cosmos.writer: A console application that can simulate telemetry data for the Trigger.

The code in this repository can be used as a complete solution or referenced as individual components.

Prerequisites

This solution requires provisioning the following resources:

  1. Azure Cosmos DB Account with a database and three containers: one for telemetry, another for the summary, and a last one for the leases (named "leases"). The containers must have the following partition keys:
    1. /id for the telemetry container.
    2. /DeviceId for the summary container.
    3. /id for the leases container.
  2. Two Azure Function Apps, one for the cosmos.trigger project and another for cosmos.monitor project, each with it's own Application Insights instance.

The Function Apps must have the following settings:

  1. cosmosConnection should either contain the connection string or have the accountEndpoint and credential set accordingly to use Managed Identities. For the latter, the Managed Identity should be provisioned and granted permissions. For more details see the trigger documentation.
  2. EVENTCONTAINER should contain the name of the container for the telemetry data.
  3. SUMMARYCONTAINER should contain the name of the container for summary data.
  4. REGION is optional and ideally contains the name of the Azure Region to leverage reading and writing data to a preferred Cosmos DB region.

Using the Change Feed Estimator

THe highlight of this repository is the cosmos.monitor TimerTrigger. It showcases the way that you can leverage the CosmosClient instance that is configured using the cosmosConnection to create a Change Feed Estimator and obtain the estimated lag on each lease, including the identifier of the owner. You can use this information to produce a dashboard and visualize or even create alerts when the lag goes beyond certain thresholds.

The following code creates an Estimator using the provided CosmosClient and referencing the Change Feed Processor configuration that the CosmosDBTrigger is wiring:

var leaseContainer = cosmosClient.GetContainer(Environment.GetEnvironmentVariable("DATABASE"), "leases");
var monitoredContainer = cosmosClient.GetContainer(Environment.GetEnvironmentVariable("DATABASE"), Environment.GetEnvironmentVariable("EVENTCONTAINER"));
this.estimator ??= monitoredContainer.GetChangeFeedEstimator("EventProcessor", leaseContainer);

It references the monitored container and the leases container, using the Function App settings value and uses a processorName that matches the CosmosDBTrigger.LeaseContainerPrefix.

The estimation results are then sent as Metrics to Application Insights with extended properties (Lease identifier and Owner):

log.LogMetric($"Estimation Lease {item.LeaseToken}", item.EstimatedLag, 
    new Dictionary<string, object>() { 
        { "Lease", item.LeaseToken }, 
        { "Owner", item.InstanceName } });

Generating traffic

The project on cosmos.writer can be used to generate a simulated load on the CosmosDBTrigger monitored container from the command-line:

dotnet run --connectionString "<cosmos DB connection string>" --concurrency 5 --devices 100000 --db <database name> --container <telemetry container name> --time PT30M

The concurrency parameter can be customized to increase or decrease the volume of concurrent operations based on the provisioned RU/s on the monitored container, while the time parameter can be used to control the duration of the simulation using ISO 8601 format.

Importing dashboard

This repository contains a sample dashboard that can be imported on Application Insights to visualize the estimated lag on the monitored container. Before importing the dashboard, you must update the Resource Ids:

"/subscriptions/{your-subscription-id}/resourceGroups/{resource-group}/providers/microsoft.insights/components/{appinsights-name}"

  1. {your-subscription-id}: The subscription Id of the Application Insights instance.
  2. {resource-group}: The resource group of the Application Insights instance.
  3. {appinsights-name}: The name of the Application Insights instance.

Dashboard