Docker Image - Expected life cycle

[george0st]

Hi @Tankilevitch, are there any rules when you create a new docker image? E.g. always after the release of the major version or after the release of the first revision for major release, etc.? Many thanks for feedback.

[Tankilevitch]

Hello, @george0st, Which Docker images are you referring to?

[george0st]

Hi @Tankilevitch, I think about MLRun (mlrun-kit) image.

[Tankilevitch]

Hi @george0st,
As part of the release process For each MLRun release we also upload all the related images to ( https://hub.docker.com/search?q=mlrun and quay.io/mlrun/<image_name>).
Currently the kit doesn't get updated to the newest MLRun version.
But there is another option ( which I'll add also to the installations docs) to use newer MLRun images.
You can update your existing release by running :

helm --namespace mlrun \
   upgrade mlrun-kit \
   --reuse-values \ 
   --set mlrun.ui.image.tag=1.0.0 \ 
   --set mlrun.api.image.tag=1.0.0 \
   --set jupyterNotebook.image.tag=1.0.0 \
   v3io-stable/mlrun-kit

Or you can remove the current one and start from the beginning, follow this to uninstall, and run the following in your helm installation step:

helm --namespace mlrun \
   install mlrun-kit \
   --set mlrun.ui.image.tag=1.0.0 \ 
   --set mlrun.api.image.tag=1.0.0 \
   --set jupyterNotebook.image.tag=1.0.0 \
   v3io-stable/mlrun-kit

I'll suggest to use only official releases and not pre-releases.

[Tankilevitch]

You can also update to newer Nuclio versions with the following configurations :

--set nuclio.controller.image.tag=<nuclio-image-version>
--set nuclio.dashboard.image.tag=<nuclio-image-version>

If I recall correctly nuclio dropped support for k8s versions lower than 1.22 from nuclio 1.7.5, so take this in mind.

[george0st]

@Tankilevitch, it is very useful information and I will use it, really many thanks.