getstackhead / module-container-docker Goto Github PK

The private registry login date set via

container:
  registries:
    - username: foo
      password: bar

is not applied. The Terraform provider file is evalated during the setup.
However the registry data is set during project deployment.

By running terraform apply periodically it should pull new containers when they are available and the "latest" tag is used.
It should then start up the new containers and update the webserver configuration (port numbers).
Then the new application should be served.

There seem to be issues, at least with private registries. After pushing a new "latest" image to the registry, terraform apply will do something, but not pull the new image? I have to run docker pull manually and then terraform apply again, to get the latest image running.

How to reproduce

I want to get the latest image of "myregistry.com:6443/myimage:latest"

docker_container

~ image = "sha256:8d8adb0c6006bb896796554e8e6220c5b394f542552097205249a743a2af319e" -> (known after apply) # forces replacement

docker_image

~ pull_triggers = [ # forces replacement
- "sha256:2dab7d4ae4052603c972c497feaaadda2ba11091f9bcd084e8600d4e0c3bfacb",
+ "sha256:f203c08336d066723f505eed25232df65d82ac9e897bc468c98aeeb16f8769d9",
]

New image hash is correct. Running docker images will yield:

myregistry.com:6443/myimage    latest       8d8adb0c6006   2 days ago     44.8MB
myregistry.com:6443/myimage <none>       5325c84777ec   2 days ago     44.8MB

Then I pull the image manually and run docker images:

docker.saitho.me:6443/simuldub_de/hugo   latest       35b55c54f1d8   10 minutes ago   45.7MB
docker.saitho.me:6443/simuldub_de/hugo   <none>       8d8adb0c6006   2 days ago       44.8MB
docker.saitho.me:6443/simuldub_de/hugo   <none>       5325c84777ec   2 days ago       44.8MB

The image is downloaded correctly. Running terraform apply again will recreate the containers and provide the latest version of the application.

It appears that when a data container, which is linked to a webserver container with volumes_from is updated and restarted, it does not affect the webserver container. The webserver will still serve the old content.

Maybe we should add a dependency to the container using the data container to the data container.

In order to have proper write permissions, the volume folders must be accessible to the user the container runs with.

• Create folders with stackhead user and group per default
• If "user" setting is defined in a service, set the user of service-specific directories to that user

StackHead is used to deploy the schemas pages schema.stackhead.io.
For some reason, running a terraform apply ALWAYS replaces the container, even if the base image has not changed.
This should be examined, as that update runs every 5 minutes.