Azure Container Instance apparently keeps running after Stop or Delete

Question

We have a microservice based app that uses Azure ServiceBus. We deploy one of the services (a saga manager) as a .Net Core console app in a docker container (Linux). We use docker-compose and a group of 2 containers (including the console app) is deployed into an Azure Container Instance.

We use MassTransit for working with Azure ServiceBus.

The console app (saga manager) starts up the service bus via MassTransit.BusControlExtensions.Start() method.

More or less at this point the console app (saga manager) is primed and ready.

Sidenote: the app in question (saga manager) runs an Automatonymous State Machine

Now if we stop or delete the container with the saga manager, and then restart it - it stops receiving bus messages.

It seems that a new instance of the saga manager (thread?) is created with each restart - while the old instance (thread?) persists somehow.

Even when we delete the Container Instance resource altogether on Azure - we can still observe that the bus messages are picked up by something...

the container's app is still alive after deleting the container group itself

Is there a way to definitively kill the detached instance/thread?

Is anyone familiar with such behavior of Azure Container Instances?

PS in the described scenarios the stop/delete operations are always successful

PS2: here is the yaml file used to deploy the problematic container group with the use of az container create command:

apiVersion: 2018-10-01
location: westeurope
name: #{groupName}#
properties:
  containers:
  - name: saga-aci
    properties:
      image: #{acrLoginServer}#/sagaazure:latest
      resources:
        requests:
          cpu: 1
          memoryInGb: 1.5
      ports:
      - port: 80
      - port: 443
      - port: 9350
  - name: proxymanager-aci
    properties:
      image: #{acrLoginServer}#/proxymanager:latest
      resources:
        requests:
          cpu: 1
          memoryInGb: 1.5
      ports:
      - port: 22999
      - port: 24000
  osType: Linux
  ipAddress:
    type: Public
    ports:
    - protocol: tcp
      port: '80'
    #- protocol: tcp
      #port: '8080'
    - protocol: tcp
      port: '22999'
    - protocol: tcp
      port: '24000'
    - protocol: tcp
      port: '443'
    - protocol: tcp
      port: '9350'
  imageRegistryCredentials:
    - server: #{acrLoginServer}#
      username: #{acrName}#
      password: #{acrPassword}#
tags: null
type: Microsoft.ContainerInstance/containerGroups

perhaps this is a restarpolicy issue?

Answer 1

For the operations of Azure Container Instance, there are some points should be paying attention to.

The stop action, if you stop the container group, then it will terminate and recycle all the containers in the group and do not preserve the containers' state. It has no effect if the container group already terminated.

The start action, if you start the container group from a stopped state, then there will be a new deployment with the same container configuration. If the image update, it will pull the new one. Start action will start the whole container group, but cannot start a special one in it.

For the delete action, I don't think the application inside the container group will still be alive if you delete the container group. The action will take a short time. But finally, the container group will be deleted totally.

You can see more details in the Azure Container Instance Operations. I have no experience about Azure ServiceBus, all above just for the Azure Container Instance. If you have any more questions please let me know.