Tracking issue for HA to beta

[fabriziopandini]

What is missing for getting HA to beta in v1.15?

• “--experimental-control-plane” flag to be renamed “--control-plane”
• “--experimental-upload-certs” flag to be renamed “--upload-certs”. Default to true?
• Improve UX for upgrade node experimental-control-plane  #1071 “kubeadm upgrade node experimental-control-plane” to be merged into “kubeadm upgrade node”
• [punted to v1.16] kubeadm join is not fault tolerant to etcd endpoint failures #1432 make join --control-plane more resilient (it should work after loosing one control plane/etcd) kubeadm: check etcd member health before trying to sync kubernetes#75600
• e2e tests
  • change kind-master job (and possibly kind-1.14) to test HA
  • change upgrade-stable-master job (and possibly upgrade-1.13-1.4) to test HA + automatic copy certificates
  • change master-on-stable job (and possibly 1.14 on 1.13) to test HA + automatic copy certificates
  • create a new job testing HA + automatic copy certificates with external etcd

[fabriziopandini]

@neolit123 opinions about the E2E part? the main idea behind is to not recreate a new series of job for HA, but instead, make current jobs have an additional control plane node so they are testing also HA

[neolit123]

“--experimental-upload-certs” flag to be renamed “--upload-certs”. Default to true?

needs discussion with the group.

“kubeadm upgrade node experimental-control-plane” to be merged into “kubeadm upgrade node”

i'd be interested in how the detection will work here.

@neolit123 opinions about the E2E part? the main idea behind is to not recreate a new series of job for HA, but instead, make current jobs have an additional control plane node so they are testing also HA

+1 on switching everything to HA.

create a new job testing HA + automatic copy certificates with external etcd

seems good. i think we can get away without upgrade for external etcd for the beta...

[neolit123]

+1 on switching everything to HA.

also i can work on these if you'd like.

[RA489]

@fabriziopandini i can work on some of them.
“--experimental-control-plane” flag to be renamed “--control-plane”
“--experimental-upload-certs” flag to be renamed “--upload-certs”. Default to true?

[neolit123]

hi @RA489 we spoke about these during yesterday's kubeadm office hours.
the idea was to do these PR last from the list from the list above.

also i think instead of renaming --experimental-control-plane -> --control-plane it might be better to have a transitional period of supporting both frags at the same time - e.g. at least 3 months.

[RA489]

@neolit123 thanks for the explanation. let me know if i can help with some others PR

[RA489]

@fabriziopandini I can look into change kind-master job (and possibly kind-1.14) to test HA

[neolit123]

create a new job testing HA + automatic copy certificates with external etcd

this is the only remaining item - PR is in flight; waits on test-grid merge approval:
kubernetes/test-infra#12906

[punted to v1.16] #1432 make join --control-plane more resilient (it should work after loosing one control plane/etcd) kubernetes/kubernetes#75600

this is a nice to have marked as priority/backlog and can be a post-beta improvement.

[fabriziopandini]

/close
As per comment above the latest open issue be a post-beta improvement and it is already tracked in #1432

[k8s-ci-robot]

@fabriziopandini: Closing this issue.

In response to this:

/close
As per comment above the latest open issue be a post-beta improvement and it is already tracked in #1432

Instructions for interacting with me using PR comments are available here. If you have questions or suggestions related to my behavior, please file an issue against the kubernetes/test-infra repository.