-
Notifications
You must be signed in to change notification settings - Fork 13
[CMS-272] Use kubeadm.k8s.io/v1alpha1 and cleanup before installation. #43
Conversation
Do we want to wrap this in some sort of logic to we can upgrade versions? |
The change is acceptable though i havent had this issue, even with new prestine deployments. Could you verify that the new apiversion works and consistently deploys? |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
lgtm
@davidewatson i think i got your verification on slack. going to merge. |
Additionally, v1alpha1 read support exists in v1.11, but will be removed in v1.12: |
@@ -63,7 +63,7 @@ | |||
|
|||
# Set up kubeadm config file to pass parameters to kubeadm init. | |||
cat << EOF | sudo tee /etc/kubernetes/kubeadm_config.yaml | |||
apiVersion: kubeadm.k8s.io/v1alpha2 |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Please add a comment line explaining when we will need to switch to v1alpha2.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
should be backwards compatible? if not yeah big deal. Good point!
@joejulian: I can add the requested logic. We'll need it later. The purpose of this PR is to ensure everyone is on the same page. I was going to submit another temporary PR along the lines of:
I think this is even worse than the current PR. :) Is it okay if we defer this change for when we rewrite these scripts? The are many problems (e.g. unpinned and untested docker releases, etc.) with these temporary scripts... |
@davidewatson I am missing context here. What are you suggestion you do with that kubeadm reset || true? I hope you are NOT suggesting we run that as part of the provisioning script. |
@alejandroEsc: That is correct. |
@davidewatson i think that brings forth a lot of assumptions that we should not make. First we assume that machines are Pristine. Making sure that is true or fixing any state issues with existing machines should not be our concern. We should make those assumptions clear and known. As for ourselves and our testing, we should do the cleaning we can and make sure things are pristine before each test cycle. That is on us. Honestly I dont want to support every possible combination of things to look for or clean, or even worry about supporting such code. |
@@ -6,6 +6,13 @@ | |||
# Run in noninteractive mode to avoid erroring when a config file preexists. | |||
export DEBIAN_FRONTEND=noninteractive | |||
|
|||
# This a temporary hack for development purposes. |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Please DO NOT INCLUDE THIS even with a TODO. Please create a new master template file and call it _DEV if you wish.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Even further, this should be in a separate PR. Note you had to use the word and
in your PR subject.
@alejandroEsc, @joejulian: I've removed the hack. Again, the intent of the PR was to simplify things while over half of our team can't seem to deploy. Let's focus on that issue please. |
No description provided.