Pro architecture
Your Adobe Commerce on cloud infrastructure Pro architecture supports multiple environments that you can use to develop, test, and launch your store.
- Master—Provides a
master
branch deployed to Platform as a service (PaaS) containers. - Integration—Provides a single
integration
branch for development, though you can create one additional branch. This allows for up to two active branches deployed to Platform as a service (PaaS) containers. - Staging—Provides a single
staging
branch deployed to dedicated Infrastructure as a service (IaaS) containers. - Production—Provides a single
production
branch deployed to dedicated Infrastructure as a service (IaaS) containers.
The following table summarizes the differences between environments:
Environment architecture
Your project is a single Git repository with three main environment branches: integration
, staging
, and production
. The following diagram shows the hierarchical relationship of Pro environments:
Master environment
On Pro projects, the master
branch provides an active PaaS environment with your production environment. Always push a copy of the production code to the master
environment so that you can debug the production environment without interrupting services.
Caveats:
-
Do not create a branch based on the
master
branch. Use the integration environment to create active branches for development. -
Do not use the
master
environment for development, UAT, or performance testing
Integration environment
The integration environment runs in a Linux container (LXC) on a grid of servers known as PaaS. Each environment includes a web server and database to test your site. See Regional IP Addresses for a list of AWS and Azure IP addresses.
Recommended use cases:
Integration environments are designed for limited testing and development before moving changes to staging and production environments. For example, you can use the integration environment to complete the following tasks:
-
Ensure that changes to continuous integration (CI) processes are Cloud compatible
-
Test critical workflows on key pages like Home, Category, Product Details Page (PDP), Checkout, and Admin
For best performance in the integration environment follow these best practices:
-
Restrict catalog size
-
Limit use to one or two concurrent users
-
Disable cron jobs and manually run as needed
Caveats:
-
Fastly CDN and New Relic services are not accessible in an integration environment
-
The integration environment architecture does not match the Staging and Production architecture
-
Do not use the
integration
environment for development testing, performance testing, or user acceptance testing (UAT) -
Do not use the
integration
environment to test B2B for Adobe Commerce functionality -
You cannot restore the database in the integration environment from the database production or staging
Staging environment
The staging environment provides a near-production environment to test your site. This environment, which is hosted on dedicated IaaS hardware, includes all services, such as Fastly CDN, New Relic APM, and search.
Recommended use cases:
The environment matches the production architecture and is designed for UAT, content staging, and final review before pushing features to the production
environment. For example, you can use the staging
environment to complete the following tasks:
-
Regression testing against production data
-
Performance testing with Fastly caching enabled
-
Test new builds instead of patching in Production
-
UAT testing for new builds
-
Test B2B for Adobe Commerce
-
Customize cron configuration and test cron jobs
See Deployment workflow and Test deployment.
Caveats:
-
After launching the production site, use the staging environment primarily to test patches for production-critical bug fixes.
-
You cannot create a branch from the
staging
branch. Instead, you push code changes from theintegration
branch to thestaging
branch.
Production environment
The production environment runs your public-facing single and multi-site storefronts. This environment runs on dedicated IaaS hardware featuring redundant, high-availability nodes for continuous access and failover protection for your customers. The production environment includes all services in the staging environment, plus the New Relic Infrastructure (NRI) service, which automatically connects with the application data and performance analytics to provide dynamic server monitoring.
Caveat:
You cannot create a branch from the production
branch. Instead, you push code changes from the staging
branch to the production
branch.
Production technology stack
The production environment has three virtual machines (VMs) behind an Elastic Load Balancer managed by an HAProxy per VM. Each VM includes the following technologies:
-
Fastly CDN—HTTP caching and CDN
-
NGINX—web server using PHP-FPM, one instance with multiple workers
-
GlusterFS—file server for managing all static file deployments and synchronization with four directory mounts:
var
pub/media
pub/static
app/etc
-
Redis—one server per VM with only one active and the other two as replicas
-
Elasticsearch—search for Adobe Commerce on cloud infrastructure 2.2 to 2.4.3-p2
-
OpenSearch—search for Adobe Commerce on cloud infrastructure 2.3.7-p3, 2.4.3-p2, 2.4.4 and later
-
Galera—database cluster with one MariaDB MySQL database per node with an auto-increment setting of three for unique IDs across every database
The following figure shows the technologies used in the production environment:
Redundant hardware
Rather than running a traditional, active-passive master
or a primary-secondary setup, Adobe Commerce on cloud infrastructure runs a redundant architecture where all three instances accept reads and writes. This architecture offers zero downtime when scaling and provides guaranteed transactional integrity.
Because of the unique, redundant hardware, Adobe can provide three gateway servers. Most external services enable you to add multiple IP addresses to an allowlist, so having more than one fixed IP address is not a problem. The three gateways map to the three servers in your production environment cluster and retain static IP addresses. It is fully redundant and highly available at every level:
- DNS
- Content Delivery Network (CDN)
- Elastic load balancer (ELB)
- Three-server cluster comprising all Adobe Commerce services, including the database and web server
Backup and disaster recovery
Adobe Commerce on cloud infrastructure uses a high-availability architecture that replicates each Pro project on three separate AWS or Azure Availability Zones, each zone with a separate data center. In addition to this redundancy, Pro staging and production environments receive regular, live backups that are designed for use in cases of catastrophic failure.
Automatic backups include persistent data from all running services, such as the MySQL database and files stored on the mounted volumes. The backups are saved to encrypted Elastic Block Storage (EBS) in the same region as the production environment. The automatic backups are not publicly accessible because they are stored in a separate system.
You can create a manual backup of the database for your Staging and Production environments using CLI commands. See Back up the database. For integration
environments, Adobe recommends creating a backup as a first step after accessing your Adobe Commerce on cloud infrastructure project and before applying any major changes. See Backup management.
Recovery Point Objective
RPO is six hours maximum time to last backup (for example at 06:00, then 12:00, then 18:00). The frequency of backups depends on the backup schedule of your plan and the volume of changes to write to the storage service.
Retention policy
Adobe retains automatic backups according to the following data retention policy:
This policy may vary depending on your cloud infrastructure plan.
Recovery Time Objective
RTO depends on the size of the storage. Large EBS volumes take more time to restore. Restoration times may vary depending on the size of your database:
- A large database (200+ GB) can take 5 hours
- A medium database (150 GB) can take 2 1/2 hours
- A small database (60 GB) can take 1 hour
Pro cluster scaling
The Pro cluster sizing and compute configurations vary depending on the chosen cloud provider (AWS, Azure), region, and service dependencies. Adobe cloud infrastructure can scale Pro clusters to accommodate traffic expectations and service requirements as demands change.
The redundant architecture empowers Adobe cloud infrastructure to upscale without downtime. When upscaling, each of the three instances rotates to upgrade capacity without impacting site operation. For example, you can add extra web servers to an existing cluster should the constriction be at the PHP level rather than the database level. This provides horizontal scaling to complement the vertical scaling provided by extra CPUs on the database level. See Scaled architecture.
If you expect a significant increase in traffic for an event or other reason, you can request a temporary increase in capacity. See How to request a temporary upsize in the Commerce Help Center.