Welcome to the Linux Foundation Forum!

LFS258 LAB12.1makes no sense (Scheduling Containers)

pipuhop
pipuhop Posts: 1
in LFS258 Class Forum

Hey There,

in Lab 12.1 is an Example for NodeSelector.

But this Example was shown with a Pod Spec with 4 Containers.
But all Containers in a Pod will always run on the same node.

You cannot influence this with a nodeselector. You can't influence that at all.

I find it quite frightening to find something like this in the Linux Foundation's course material.

Or have I got something completely wrong here?

Thanks.

Comments

  • admusin
    admusin Posts: 7

    Hi pipuhop,

    I find the wording of this lab kind of confusing too (just went through it), as some paragraphs sort of can be interpreted (depending how you read them) that containers of the pod "could" be spread across nodes.

    That would contradict what earlier chapters of the course were stating about the pod being the "unit of scheduling".

    I find a bit strange also the choice of using crictl to simply inspect where the pod was deployed (along with its containers) when usually in day-to-day life a K8s admin would simply use kubectl get pods -o wide to find that out much faster, without relying on the crictl config being already set or having to provide that long --runtime-endpoint parameter at a minimum, if it is not.

Categories

Upcoming Training