Documentation Center

Deployment scenarios for Docker images

SDL Machine Translation Edge supports several scenarios for Docker image deployment.

Preserve Job Queue across Master container restarts

By default, the state of the SDL Machine Translation Edge cluster is stored inside the Master container file system. If the container is removed, the SDL Machine Translation Edge Job Queue is lost. This behavior may be desirable or not, depending on the use-case.

In order to make the SDL Machine Translation Edge Queue persistent, the directory /opt/sdl/ets/data can be mounted from an external volume, thus it will persist when the Master container is restarted. The mounting procedure is specific to the orchestration used to start the container (Docker Swarm, Kubernetes, etc.)

Store Language Pairs inside the containers vs. Mount Language Pairs from a volume outside the container

Depending on the constraints regarding container size, the Language Pairs could be installed:
  • Inside the container image with the following alternatives:
    • The entire cluster uses the same image.
    • Worker containers which host Translation Engines could use different images.
  • On a volume outside the container

SDL Machine Translation Edge will use any Language Pair as long as the Language Pair is present in the container file system and SDL_ETS_LP_INSTALL_PATH points to it.

Installing the Language Pairs inside the containers tends to be simpler to deploy, with the minor drawback of having larger sized Docker images.