As the application is already running as set of containers or pods, it can be scaled by increasing the replica of running instance.
Except for scheduler, everything can have multiple replicas.
Scale MariaDB in case of DB performance required. Add BI tools to Read Replicas.
Scale Gunicorn in case of API performance required.
Scale Workers in case background jobs queue performance is required.
In case of Kubernetes, use official helm chart replace with your custom values.yaml. In case helm is not used in your infrastructure then generate a deploy.yaml template using helm template command. Setup Horizontal Pod Autoscaler for on standard deployments and custom worker deployments as per the load.
In case of other orchestrators the base requirements are:
Use Filesystem which supports symlinks to mount sites directory. Static assets under sites/assets have internal symlinks.
All pods need to connect to mariadb and redis services.
Run separate Jobs or Tasks for setting common_site_config.json and creating new site.