Share via

Service fabric service preventing Docker service start

nedukull 1 Reputation point
2026-04-06T13:06:59.94+00:00

I have custom image where docker CE already installed. Docker is up and running. When I try to spin off Azure Service Fabric using same custom image. Docker service is stopped. Still all the docker functionalities like ,pull image, run container working as expected. When I stop SVCFabric Services and start the docker it started. The moment when I started Service Fabric again docker going to STOPPED state.

OS: Windows Data Center 2025

Docker : Docker CE Latest

Azure Service Fabric
Azure Service Fabric

An Azure service that is used to develop microservices and orchestrate containers on Windows and Linux.


2 answers

Sort by: Most helpful
  1. Ankit Yadav 13,600 Reputation points Microsoft External Staff Moderator
    2026-04-06T13:20:03.3133333+00:00

    Hello Nedukull,

    This behavior is expected on Azure Service Fabric clusters running on Windows when Docker CE is installed.

    On Windows, Service Fabric does not treat Docker as an independent, user‑managed service. Instead, Service Fabric expects to own and manage the container runtime lifecycle itself. When the Service Fabric runtime starts, it attempts to start and control the container engine using its own configuration. If a separately installed and managed Docker CE service is present, Service Fabric can stop or override that service, which is why you see the Docker service go to a Stopped state as soon as Service Fabric starts.

    Let me know if that answers your query!

    0 comments No comments

  2. Q&A Assist
    2026-04-06T13:07:12.9633333+00:00
    AI generated content. This question contains an answer created with AI from Q&A Assist. Learn more

    Service Fabric on Windows does not support Docker CE and expects to own and manage the container runtime service itself. On supported Windows versions, Service Fabric is designed to work with Mirantis Container Runtime (or Docker EE on older Windows Server versions), not Docker CE. When Service Fabric starts, it manages the Docker/engine service according to its own configuration, which is why the Docker service is being stopped as soon as Service Fabric comes up.

    Key points from the supported model:

    1. Supported container runtimes on Windows
      • Windows Server 2022: Mirantis Container Runtime
      • Windows Server 2019/2016: Docker EE Docker CE is not listed as a supported runtime for Service Fabric on Windows, so behavior like the engine service being stopped or overridden by Service Fabric is expected and unsupported.
    2. How Service Fabric interacts with the Docker daemon
      • Service Fabric starts the Docker daemon with specific arguments defined in the cluster manifest under the Hosting section.
      • When custom arguments are specified via ContainerServiceArguments, Service Fabric does not pass any other arguments to the Docker engine except --pidfile, and it expects the daemon to listen on the default named pipe.
      • This means Service Fabric assumes control of how the Docker/engine service is started and can conflict with a pre‑installed and separately managed Docker CE service.
    3. Recommended approach
      • Use a supported container runtime for Service Fabric on Windows (Mirantis Container Runtime on Windows Server 2022, or Docker EE on 2019/2016) instead of Docker CE.
      • Let Service Fabric manage the container runtime service rather than trying to keep a separate Docker CE service running in parallel.
      • Configure any custom daemon arguments via the Service Fabric cluster manifest (ContainerServiceArguments in the Hosting fabricSettings) instead of directly managing the Docker service.

    Given the current setup (Windows Datacenter 2025 + Docker CE), the conflict where Service Fabric stops the Docker service is a consequence of running an unsupported combination. Moving to a supported runtime and allowing Service Fabric to manage it is required for a stable configuration.


    References:

Your answer

Answers can be marked as 'Accepted' by the question author and 'Recommended' by moderators, which helps users know the answer solved the author's problem.