Deploy and Manage a Scalable Moodle Cluster on Azure

This repo contains guides and Azure Resource Manager templates designed to help you deploy and manage a highly available and scalable Moodle cluster on Azure. In addition, the repo contains other useful information relevant to running Moodle on Azure such as a listing of Azure-relevant Moodle plugins and information on how to offer Moodle as a Managed Application on the Azure Marketplace or on an IT Service Catalog.

If you have Azure account you can deploy Moodle via the Azure portal using the button below, or you can deploy Moodle via the CLI. Please note that while you can use an Azure free account to get started depending on which template configuration you choose you will likely be required to upgrade to a paid account.

Fully configurable deployment

The following button will allow you to specify various configurations for your Moodle cluster deployment. The number of configuration options might be overwhelming, so some pre-defined/restricted deployment options for typical Moodle scenarios follow this.

NOTE: All of the deployment options require you to provide a valid SSH protocol 2 (SSH-2) RSA public-private key pairs with a minimum length of 2048 bits. Other key formats such as ED25519 and ECDSA are not supported. If you are unfamiliar with SSH then you should read this article which will explain how to generate a key using the Windows Subsystem for Linux (it's easy and takes only a few minutes). If you are new to SSH, remember SSH is a key pair solution. What this means is you have a public key and a private key, and the one you will be using to deploy your template is the public key.

Predefined deployment options

Below are a list of pre-defined/restricted deployment options based on typical deployment scenarios (i.e. dev/test, production etc.) All configurations are fixed and you just need to pass your ssh public key to the template for logging in to the deployed VMs. Please note that the actual cost will be bigger with potentially autoscaled VMs, backups and network cost.

Deployment Type	Description	Launch
Minimal	This deployment will use NFS, Microsoft SQL, and smaller autoscale web frontend VM sku (1 core) that'll give faster deployment time (less than 30 minutes) and requires only 2 VM cores currently that'll fit even in a free trial Azure subscription.
Small to Mid-Size	Supporting up to 1000 concurrent users. This deployment will use NFS (no high availability) and MySQL (8 vCores), without other options like elastic search or redis cache.
Large size deployment (with high availability)	Supporting more than 2000 concurrent users. This deployment will use Gluster (for high availability, requiring 2 VMs), MySQL (16 vCores) and redis cache, without other options like elastic search.
Maximum	This maximal deployment will use Gluster (for high availability, adding 2 VMs for a Gluster cluster), MySQL with highest SKU, redis cache, elastic search (3 VMs), and pretty large storage sizes (both data disks and DB).

NOTE: Depending on the region you choose to deploy the stack in - the deployment might fail due to SKUs being hardcoded in the template where they are not available. For example, today our small-mid-size deployment option hard codes Gen-4 Azure MySQL SKUs into the template, and if a region where that is currently not available in (i.e. westus2) is used, your deployment will fail. If your deployment fails, please revert to the fully configurable template where possible and change the SKU paramater to one that exists in your region (i.e. Gen-5) or alternatively change your deployment region to one in which the SKU is available (i.e. southcentralus).

Stack Architecture

This template set deploys the following infrastructure core to your Moodle instance:

• Autoscaling web frontend layer (Nginx for https termination, Varnish for caching, Apache/php or nginx/php-fpm)
• Private virtual network for frontend instances
• Controller instance running cron and handling syslog for the autoscaled site
• Azure Load balancer to balance across the autoscaled instances
• Azure Database for MySQL or Azure Database for PostgreSQL or Azure SQL Database
• Dual GlusterFS nodes or NFS for high availability access to Moodle files

This template set optionally configures the following additional infrastructure:

• Azure Backup for Moodle site backups
• Azure Blob Storage for ObjectFS (Moodle sitedata)
• Azure Application Gateway for SSL offloading and WAF
• Azure Redis Cache instance for Moodle caching
• Azure DDoS Protection plan to secure your Moodle site from DDoS attacks
• Azure Key Vault for storing your CA Cert for your Moodle site
• Azure Search instance or three Elasticsearch VMs for HA Global Search in Moodle
• Apache Tika VMs for search indexing in Moodle

The template also optionally installs plugins that allow Moodle to be integrated with select Azure services (see below for details).

Useful Moodle plugins for integrating Moodle with Azure Services

There below is a listing of useful plugins allow Moodle to be integrated with select Azure services:

• Object File System Plugin* for Azure Blob Storage
• Azure Search Plugin* for Azure Search
• Trigger Plugin and Restful Webservice Plugin for Azure Logic Apps (requires use of Moodle Connector now in development)
• Office 365 and Azure Active Directory Plugins for Moodle* for Azure Active Directory
• Elasticsearch Plugin*

At the current time this template allows the optional installation of all of the plugins above with an * next to them. Please note these plugins can be installed at any time post deployment via Moodle's own plugin directory. You can find a list of all Azure relevant plugins in the Moodle plugin directory here. You might also choose to follow this list via RSS.

Moodle as a Managed Application

You can learn more about how you can offer Moodle as a Managed Application on the Azure Marketplace or on an IT Service Catalog here. This is a great read if you are offering Moodle hosting services today for your customers.

Observations about the current template

The template is highly configurable, full details of the configuration options can be found in our documentation (more specifically in our parameters documentation). The following sections describe observations about the template that you will likely want to review before deploying:

Scalability Out system is designed to be highly scalable, to achieve this we provide a Virtual Machine Scaleset for the web tier. This is already configured to scale on high load. However, scalaing the VMs is not instantaneous, therefore if you have a known hihg-load situation at a given time (e.g. exam) you should manually scale in preparation. This can be done through the Azure portal or the CLI. The database is less easily scaled at this point, but it is possible and documented in our management documentation.

SSL The template fully supports SSL but it is not possible for the template to manage this for you. More information in our managing certs documentation.

Moodle PHP Code The Moodle PHP code is stored on the Controller VM and copied to each front end VM upon deployment and upon request (should you update the Moodle code with your own code). For more information see our management documentation.

Database Currently the best performance is achieved with Azure Database for MySQL and Azure SQL Database. With Azure Database for PostgreSQL we have hit database constraints which caused processes to load up on the frontends until they ran out of memory. It is possible some PostgreSQL tuning might help here. At this stage Azure Database for MySQL and PostgreSQL do not support being moved to a vnet. As a workaround, we use a firewall-based IP restriction allow access only to the controller VM and VMSS load-balancer IPs.

File Storage There are two options for file storage (moodledata) - Gluster FS and NFS. The Gluster FS solution is replicated thus provides highler availability, but incurs additional cost (2 x VMs) and some performance penalties (we are exploring ways to improve this and would welcome contributions from people who know Moodle and/or Gluster). NFS is highly performant and utilizes an existing VM in the cluster (so lower cost), but it is a single point of failure. At the time of writing there is no simple way to switch from one to the other depending on expected workloads and availability requirements, again this is something we would love to see resolved.

Search. Azure supports running an Elasticsearch cluster, however it does not offer a fully-managed Elasticsearch service, so for those looking for a fully-managed Search service Azure Search is recommended.

Caching. While enabling Redis cache can improve performance for a large Moodle site we have not seen it be very effective for small-to-medium size sites. We can likely improve upon this, patches welcome ;-)

Regions. Note that not all resources types (such as databases) may be available in your region. You should check the list of Azure Products by Region to for local availability.

Common questions about this Template

1. Is this template Moodle as IaaS or PaaS? While the current template leverages PaaS services such as Redis, MySQL, Postgres, MS SQL etc. the current template offers Moodle as IaaS. Given limitations to Moodle our focus is IaaS for the time being however we would love to be informed of your experience running Moodle as PaaS on Azure (i.e. using Azure Container Service or Azure App Service).

2. The current template uses Ubuntu. Will other Operating Systems such as CentOS or Windows Server be supported in the future? Unfortunately we only have plans to support Ubuntu at this time. It is highly unlikely that this will change.

3. What configuration do you recommend for my Moodle site? The answer is it depends. At this stage we provide some rudimenatary t-shirt sized deployment recommendations and we are still building out our load testing tools and methodologies to provide more granularity. With that being said this is an area we are investing heavily in this area and we would love your contributions (i.e. load testing scripts, tools, methodologies etc.).

If you have an immediate need for guidance for a larger sized deployment, you might want to share some details around your deployment on our issues page and we will do our best to respond. Please as much information about your deployment as possible such as:

• average number of concurrent users your site will see
• maximum level of concurrent/simultaenous users your site needs to support
• whether or not HA is needed
• any other attributes specific to your deployment (i.e. load balancing across regions etc.)

4. Did Microsoft build this template alone or with the help of the Moodle community? We did not build this template alone. We relied on the expertise and guidance of many capable Moodle partners around the world. The initial implementation of the template was done by Catalyst IT.

5. How does this template relate to other Moodle offerings available on the Azure Marketplace? It is generally not a good idea to run Moodle as a single VM in a production setting. This template is highly configurable and allows for high availability and redundancy.

6. How does this template relate to this Azure Quickstart Template for Moodle? This repo is the working repo for the quickstart template. We will be pushing changes from this template to the quickstart template on a regular cadence.

7. I am already running Moodle on Azure. How does this work benefit me? We are looking for painpoints from you and the broader Moodle on Azure community that we can help solve. We are also looking to understand where our implementation of Moodle on Azure outperforms or underperforms other implementations such as yours that are out in the wild. If you have observations, performance benchmarks or just general feedback about your experience running Moodle on Azure that you'd like to share we're extremely interested! Load testing is a very big area of focus, so if you have scripts you wouldn't mind contributing please let us know.

8. Has anyone run this template sucessfully in production? Yes they have. With that being said we do not make any performance guarantees about this architecture.

9. What type of improvements have you succeeded in making Since we first began this effort we have managed to make great gains, achieving a >2x performance boost from our original configuration by making tweaks to things like where PHP files were stored. Our work is nowhere near over.

10. What other Azure services (i.e. Azure CDN, Azure Media Services, Azure Bot Service etc.) will you be integrating with when this effort is complete? It's not clear yet. We'll need your feedback to decide.

11. Why the database on a public subnet? At this stage Azure Database for MySQL and PostgreSQL do not support being moved to a vnet. As a workaround, we use a firewall-based IP restriction allow access only to the controller VM and VMSS load-balancer IPs.

12. How can I help with this effort? Please see below.

Automated Testing (Travis CI)

This repository uses Travis CI to deliver automated testing.

The following tests are carried out for every Pull Request and will also run in a Travis CI enabled forked repository:

• JSON Linting - All JSON files are linted to ensure they do not contain any syntax errors.
• JSON Code Style - All JSON files are tested to ensure they comply with project code style rules.

The following tests are carried out as part of the Pull Request merging prior to a contribution being accepted into the release branch:

• Template Validation - The template is subbmitted to Azure to ensure it is correclty formatted and contains valid logic.
• Template Build - The template is submitted to Azure and the stack described in the template is built to ensure a stack is correctly deployed.

Setting Up Travis CI for Template Build

The following describes the process required if you want to run the template validation and build steps using your own Travis and Azure accounts.

To set up the build process, you will need:

• An Azure account or active subscription
• A fork of this repository linked to Travis CI
• Access to an installed instance of the Azure CLI
• A SSH keypair

The Travis CI process uses the Azure CLI Service Principal login method to authenticate against Azure. The documentation for logging in via a Service Principal can be found here: https://docs.microsoft.com/en-us/cli/azure/authenticate-azure-cli?view=azure-cli-latest#logging-in-with-a-service-principal

Before you can log in using the Service Principal process you need to create a Service Principal. The documentation to create a Service Principal login can be found here: https://docs.microsoft.com/en-us/cli/azure/create-an-azure-service-principal-azure-cli?view=azure-cli-latest

When a Service Principal is created using the Azure CLI a JSON response is returned containing:

• name - This is the Service Principal username.
• password - This is the Service Principal password.
• tenantId - This is the Service Principal tenant unique ID.

You will need these three above values to have Travis and Azure deploy and test your template.

The next step is to take the above values returned by the Service Principal creation and use them to define environment variables in Travis CI.

The following link shows how to set up per repository environment variables in Travis CI: https://docs.travis-ci.com/user/environment-variables/#Defining-Variables-in-Repository-Settings Using this documention set up the following three hidden environment variables in Travis CI for your fork of this repository.

• SPNAME - The value of the name parameter returned by the Service Principal create proccess.
• SPPASSWORD - The value of the password parameter returned by the Service Principal create proccess.
• SPTENANT - The value of the tenant parameter returned by the Service Principal create proccess.
• SPSSHKEY (default: generate new)- A public SSH key that you have the corresponding private key for. This is currently not used but is required for the build to be successful.
• LOCATION (default: southcentralus)- Location for the test resource group.
• RESOURCEGROUP (default: azmdl-travis-XXX)- Name to use for the resource group.
• FULLCI_BRANCHES (default: master)- Name of branches (separated by ':') to always run FULL CI (if credentials are provided). Full CI will run a deployment test which will create and use resources from your Azure account.

NOTE: You can trigger a full CI test by adding [full ci] or [fullci] anywhere in the commit message.

NOTE: Make sure you set the environment variables to hidden otherwise they will be exposed publically at run time.

NOTE: As per the Travis CI documentation make sure you have correctly escaped the enviroment variable values when they are defined.

Once the environment variables are defined, Travis CI will run the template validate and build steps as part of the test process.

Contributing

This project welcomes contributions and suggestions. Our goal is to work on Azure specific tooling for deploying and managing the open source Moodle learning management system on Azure. We do not work on Moodle itself here, instead we work upstream as appropriate.

The short version of how to contribute to this project is "just do it". Where "it" can be defined as any valuable contribution (and to be clear, asking questions is a valuable contribution):

• ask questions
• provide feedback
• write or update documentation
• help new users
• recommend the project to others
• test the code and report bugs
• fix bugs and issue pull requests
• give us feedback on required features
• write and update the software
• create artwork
• translate to different languages
• anything you can see that needs doing

For a more detailed discussion of how to contribute see our Contribution Guide.

Code of Conduct

This project has adopted the Microsoft Open Source Code of Conduct. For more information see the Code of Conduct FAQ or contact opencode@microsoft.com with any additional questions or comments.

Legal Notices

Microsoft and any contributors grant you a license to the Microsoft documentation and other content in this repository under the Creative Commons Attribution 4.0 International Public License, see the LICENSE file, and grant you a license to any code in the repository under the MIT License, see the LICENSE-CODE file.

Microsoft, Windows, Microsoft Azure and/or other Microsoft products and services referenced in the documentation may be either trademarks or registered trademarks of Microsoft in the United States and/or other countries. The licenses for this project do not grant you rights to use any Microsoft names, logos, or trademarks. Microsoft's general trademark guidelines can be found at http://go.microsoft.com/fwlink/?LinkID=254653.

Privacy information can be found at https://privacy.microsoft.com/en-us/

Microsoft and any contributors reserve all others rights, whether under their respective copyrights, patents, or trademarks, whether by implication, estoppel or otherwise.

Next Steps

1. Deploy a Moodle Cluster
2. Obtain Deployment Details about a Moodle Cluster
3. Delete a Moodle Cluster