Hardware requirements
See the hardware requirements for the requirements of each application and the services.
Applications
The server and delivery are both applications written in Node.js. Node.js is single threaded, but it can handle concurrency through the asynchronous event loop. If a Node.js process crashes, it has to be restarted and it is not able to accept requests during startup.
For a production setup, Node.js processes should always be redundant to prevent downtime in case of crashes.
Docker
We recommend to use Docker as we provide Dockerfiles for every application and service. Compatibility is ensured with every release.
Services
Both Elasticsearch and Postgres will need a persistent volume mounted. Any custom configuration, as well as the supported versions for Elasticsearch and Postgres are visible in the respective Dockerfile.
- https://github.com/livingdocsIO/dockerfile-elasticsearch
- https://github.com/livingdocsIO/dockerfile-postgres
Applications
The applications are stateless and follow the 12 factor app methodology. Any system level dependency and the required environment variables are visible in the respective Dockerfile.
- https://github.com/livingdocsIO/docker/tree/master/livingdocs-server-base
- https://github.com/livingdocsIO/docker/tree/master/livingdocs-editor-base
Container data volumes
Both the containers for server and editor are stateless. The Elasticsearch and Postgres containers data directory needs to be mounted to the host on with a data volume.
Health checks
- Editor: HTTP GET /status, Port 9000
- Server: HTTP GET /status, Port 9090
- Postgres: TCP, Port 5432
- Elasticsearch: TCP, Port 9200
- Redis: TCP, Port 6379
Deployment
We recommend building Docker images on CI and pushing them to the registry. Deployment can be done manually or triggered by CI continuously.
Operating Livingdocs
Notable required configurations and our recommended best practices are described below.
Avoiding CORS
The preferred solution is to serve the Livingdocs Server instance on the same domain as the editor to prevent CORS requests and have a better security as we don’t need to make the login cookies accessible on multiple domains. Please expose the Livingdocs Server instance on /proxy/api and then in the livingdocs-editor environment configs, configure the host: module.exports={api: {host: '/proxy/api'}}
Alternatively you can use proxiedHost instead of host to proxy to a dns name of an internal service that is not accessible from the internet. We will automatically set up a http/websocket proxy on /proxy/api.
