Greengress IoT Deployment while deploying docker container from ECR doesn't stop previous run

Question

I have my docker container hosted in ECR, my component recipe is:

{
  "RecipeFormatVersion": "2020-01-25",
  "ComponentName": <component-name>,
  "ComponentVersion": "1.0.46",
  "ComponentType": "aws.greengrass.generic",
  "ComponentDescription": "A component that runs a Docker container from a private Amazon ECR image.",
  "ComponentPublisher": "Amazon",
  "ComponentDependencies": {
    "aws.greengrass.DockerApplicationManager": {
      "VersionRequirement": ">=2.0.0 <2.1.0",
      "DependencyType": "HARD"
    },
    "aws.greengrass.TokenExchangeService": {
      "VersionRequirement": ">=2.0.0 <2.1.0",
      "DependencyType": "HARD"
    }
  },
  "Manifests": [
    {
      "Platform": {
        "os": "all"
      },
      "Lifecycle": {
        "Run": "docker run --rm --net=host <image_name>",
        "Shutdown": "docker stop $(docker ps -a -q --filter ancestor=<image_name> --format=\"{{.ID}}\")"
      },
      "Artifacts": [
        {
          "Uri": <image_uri>,
          "Unarchive": "NONE",
          "Permission": {
            "Read": "OWNER",
            "Execute": "NONE"
          }
        }
      ]
    }
  ],
  "Lifecycle": {}
}

My question is, sometimes when the internet connection is slow, the previous deployment's docker run doesn't stop, so I end up having 2 running docker containers. How can I ensure the previous one stops, I have added -rm and a SHUTDOWN command.

Gavin_A · Answer

Hi!

You're on the right approach. In the recipe file above, you are mixing `Run` with `Shutdown`. `Run` stands on its own and is good for foreground processes. But in that case, you must honor and process the SIGTERM received.

With `Shutdown` you would use `Startup` instead of `Run`. For that, `Startup` would do a `docker run -d --name CONTAINER_NAME` and name the container, while `Shutdown` would do a `docker stop CONTAINER_NAME`.

In summary, `Run` requires you to run the component in foreground and process SIGTERMs when Greengrass issues a stop to the component. `Startup` and `Shutdown` require a exit code of 0 for each, and your Shutdown process would need to know the running container (or process).

HTH!

JoeAtAWS · Answer

Hi, can you try forwarding signals to the container itself?

Add `--sig-proxy=true` to your run command. You can remove the shutdown phase.

Greengress IoT Deployment while deploying docker container from ECR doesn't stop previous run

Add your answer

Relevant content