-
Notifications
You must be signed in to change notification settings - Fork 39.4k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Update kubeadm etcd to 3.0.13 in order to switch to the etcd3 storage format #35921
Conversation
Jenkins GKE smoke e2e failed for commit d62609e1da5ff33b2cd11fc9d486a4d2ae95c7e6. Full PR test history. The magic incantation to run this job again is |
lgtm. |
If @timothysc says it's good, I'm happy. |
/lgtm |
Jenkins unit/integration failed for commit d62609e1da5ff33b2cd11fc9d486a4d2ae95c7e6. Full PR test history. The magic incantation to run this job again is |
Jenkins Kubemark GCE e2e failed for commit d62609e1da5ff33b2cd11fc9d486a4d2ae95c7e6. Full PR test history. The magic incantation to run this job again is |
It flaked and I'll retrigger soon, but waiting on @wojtek-t to actually push the image... |
IIRC, we said that 2.x is what we should stick to for the moment, but please correct me if I'm wrong. If that is correct, we should probably close this PR. Also, something to consider, we could provide a way for folks to test with 3.0 (i.e. via a version flag, which would probably have to go in the config file). |
We just said that we'll wait until the 3.0.13 or 3.0.14 image is pushed == until the k8s etcd team consider etcd3 stable, which will be in time for v1.5. So we'll just defer until they have pushed the image and then merge this in time for v1.5 kubeadm But yes, we should expose this version in the api as well (I've started working on a such change) |
@wojtek-t @luxas @errordeveloper : Is this related to the exception being requested in: |
d62609e
to
aecb099
Compare
Ok, so a The reason is that GKE won't upgrade to etcd3 for v1.5, so then no final 3.0.14 image should exist. |
The decision was that etcd3 wouldn't be the default in 1.5 for any deployments, not just GKE. |
Ah got it. Anyway, since we're only starting new clusters for the time being, it's okay with etcd3. @errordeveloper @timothysc Want to give your lgtm? |
@mikedanese PTAL and LGTM, this is really straightforward and has been lgtm'd once already |
So etcd3 was booted from the release, what's the reason for pushing this through? |
The reason it didn't make v1.5 was upgrades from existing clusters. It was reported that etcd3 worked well when running it in new clusters. And if we wait with the etcd upgrade until kubeadm is in beta, we have to deal with the etcd2 => etcd3 upgrade which we can avoid by upgrading now. |
Is kubeadm still alpha? Then I'm ok with it. |
Yes, it's alpha |
/lgtm |
Automatic merge from submit-queue |
I thought that 3.0.x did not compile/work on ARM kit eg RPI - 3 There is a etcd2 compiled version of etcd but I could not find a etc3 version. As I understand it this will support ARM this could make that difficult? |
ref: #35723
I think we should switch as soon as possible, but run it in etcd2 mode until the full etcd3 mode is stable
@kubernetes/sig-cluster-lifecycle @wojtek-t @xiang90 @lavalamp
This change is