Deploy a Meilisearch instance on Azure

The following is a guest post by Christopher Maneu, Data Engineering Lead Advocate at Microsoft.

In this tutorial, we will see how to deploy a Meilisearch instance on Azure, and also some considerations for hosting a production Meilisearch instance on Azure.

You can deploy a Meilisearch instance via the official Meilisearch Docker image by clicking the following button:

.

When clicking this button, you'll be redirected to the Azure Portal and asked a few questions:

• "Region": This is the deployment region. You should select a region close to you or your users
• "Environment": The name of the environment you want to deploy to, like dev (for development) or prod (for production). This setting has no effect on the type of resources deployed. It's just for convenience
• "Application Name": The name of your application. This name should be unique across all Azure customers. The URL where your application is deployed will contain this value
• "Location": This is the resources' location. For reasons out of scope of this documentation, this parameter is different from the Region one. By default, it is set to eastus, but you can type any Azure location (like francecentral, westeurope, or japaneast)
• "Meilisearch_apikey": This will be the master key of your Meilisearch instance. While we generate a random key by default, we encourage you to set your own key

Once you have filled in these fields, click the "Review + create" button and then the "Create" button.

After a few minutes, the deployment will be complete. You'll be able to access your instance URL by clicking on the "Outputs" tabs on the left.

This button uses an Infrastructure as Code bicep file. The code is open-source and hosted on GitHub. You can submit PRs if needed, or use it in your own projects.

This code will set up two main components:

• An Azure App Service instance to host the Meilisearch container
• An Azure File Share which is volume-mapped into the Meilisearch container for database persistence

We've made some opinionated choices in the one-click deployment that should work for most production-ready deployments. This section explains some of the choices you need to be aware of in the context of a production deployment.

The one-click button creates a storage with the "LRS" redundancy option. That means that your data is replicated across multiple servers, but in the same data center. You can change this setting to choose a multi-zone or a multi-region redundancy option by modifying the template.

By default, there is no automatic backup of your index. If you want to use Azure Backup capabilities to achieve this result, you can look at Azure File Share Backup.

To enable free tier, the Always-on option is disabled by default. That means that after some time without any traffic, your instance will be deactivated. When the next client hits your instance, it'll be started again automatically. However, you will have a response time in the dozen-seconds range.

You can avoid this cold start by enabling the Always-on feature, either in the Azure portal or in the Infrastructure as Code.

Azure App Service offers a horizontal auto-scale feature that allows you to automatically add more servers to process incoming requests.

At time of writing, Meilisearch is not ready to handle multiple servers using the same underlying storage. Enabling this feature can cause unexpected results and should not be considered for a production environment.

If you need more processing power, you can use the Scale-Up feature of App Service.