Skip to content

[pull] master from kubernetes:master - #1924

Open
pull[bot] wants to merge 9026 commits into
next-stack:masterfrom
kubernetes:master
Open

pull[bot] wants to merge 9026 commits into
next-stack:masterfrom
kubernetes:master

Conversation

@pull

@pull pull Bot commented Oct 24, 2021 •

Copy link
Copy Markdown

See Commits and Changes for more details.


Created by pull[bot]

Can you help keep this open source service alive? 💖 Please sponsor : )

@pull pull Bot added the ⤵️ pull label Oct 24, 2021
upodroid and others added 29 commits September 13, 2026 10:48
build kubetest2 binaries and publish them to the staging bucket
Add availability zones for ap-east-2 (Taipei), ap-southeast-5
(Malaysia), ap-southeast-6 (New Zealand), ap-southeast-7 (Thailand)
and mx-central-1 (Mexico) to the well-known AWS zone list used for
shell completion of `kops create cluster --zones`.

kOps already works in these regions because the region is validated
against ec2:DescribeRegions at runtime; only the completion list was
out of date.
Got the IG creation working.
Kubelet is coming up on the etcd IG.
Currently having issues with getting the etcd to attach to peers.
fixed goimports
Fixed ptr to to new on count.
Fixed image rewrite to work for all of control plane.
Fixed docs to reflect changes.
Fixed role direct equality check.
Fix node check for bastion.
Fix KubControllerManager to KubeControllerManager.
Fixed experimental roles test.
Fixed nodelabel tests.
Fix label test.
The v1alpha2 fuzzer cleared CloudProvider.AWS.BinariesLocation and
CloudProvider.GCE.BinariesLocation because v1alpha2 had nowhere to put
them. #18740 gave them a home: CloudConfiguration.AWSBinariesLocation
and GCEBinariesLocation, written in Convert_kops_ClusterSpec_To_
v1alpha2_ClusterSpec and read back in the reverse direction.

The fixup outlived that, so the field the round trip was meant to cover
is cleared before it is ever exercised. Dropping it restores the
coverage for a field that has already been lost in conversion once.

TestRoundTripTypes passes over 50 seeds without the fixup.

Signed-off-by: SaiPisey2 <piseysai0202@gmail.com>
zones: add newer AWS regions to --zones completion
Commented out gcpmachine check
Make control plane split by role work.
aws: Add Amazon ECR dual-stack endpoints for IPv6 support
- Add Kubernetes 1.37 to the alpha and stable channels (1.37.1).
- Bump the alpha channel to 1.36.5, 1.35.9 and 1.34.12.
- Add asset hashes for v1.34.12, v1.35.9, v1.36.5, v1.37.0 and v1.37.1.
Upstream no longer publishes releases to Docker Hub (the last one there
is v0.6.7), so pull the image from ECR Public.

The update brings newer Go and dependency versions, and support for
ConfigMap entries with a "*" namespace, which kOps generates for
serviceAccountExternalPermissions that use namespace "*".

admissionReviewVersions stays at v1beta1: v0.6.17 still returns
AdmissionReview responses without apiVersion and kind, which the API
server rejects for v1.
chore(channels): add Kubernetes 1.37 and September patch releases
aws: update pod-identity-webhook to v0.6.17
kubernetes-prow Bot and others added 30 commits October 25, 2026 17:22
feat(containerd): add env to the containerd spec
Re-derive the manifest from upstream deploy/ at v0.6.19, which adds the
eks.amazonaws.com/skip-pod-identity-webhook objectSelector entry.

admissionReviewVersions stays at v1beta1: switching to v1 together with
the image would reject responses from the old v0.6.17 pods during the
rollout.
aws: update pod-identity-webhook to v0.6.19
Reject unsupported providers instead of rendering an AWS NodeClass reference.
parameterize the Karpenter NodeClass group and kind
kops-controller imports pkg/commands for ConfigBuilder, so this CLI-only file pulled kubectl, cli-runtime and kustomize into the controller. Moving it next to the other kops commands removes 119 packages and about 55 MB from the kops-controller binary.
refactor: move the helpers command from pkg/commands to cmd/kops
bare-metal e2e: bound apt-get update and artifact ssh connect time
chore(channels): bump alpha to October patches
Since kOps 1.36, nodeStatusUpdateFrequency is written to the kubelet
configuration file instead of being passed as a flag. When that field
is set in the file and nodeStatusReportFrequency is not, the kubelet
defaults nodeStatusReportFrequency to the same value instead of 5m.
A cluster that sets nodeStatusUpdateFrequency to a few seconds now has
every kubelet post its full node status at that period, even when
nothing changed, which multiplies Node writes and watch traffic.

Add nodeStatusReportFrequency to the kubelet spec and map it into the
generated KubeletConfiguration, so such clusters can restore the 5m
report period. When unset, the kubelet defaults apply as before.
fix(assets): reject conflicting hashes for the same file download URL
Add Packages (*PackagesConfig) to KubeletConfigSpec to allow overriding the kubelet binary URL and hash per architecture on a cluster-wide or per-InstanceGroup basis.

Signed-off-by: Arnaud Meukam <ameukam@gmail.com>
feat(kubelet): support custom kubelet binary package
chore(channels): promote October patches to stable
Signed-off-by: Ciprian Hacman <ciprian@hakman.dev>
Signed-off-by: Ciprian Hacman <ciprian@hakman.dev>
…ort-frequency

feat(kubelet): add nodeStatusReportFrequency to the kubelet spec
The SDK retries CreateBucket. When the first attempt succeeded, the retry gets 409 BucketAlreadyOwnedByYou and Up failed (ci-kubernetes-e2e-ec2-scale-correctness-100, 2026-10-24).

Signed-off-by: Davanum Srinivas <davanum@gmail.com>
kubetest2-kops: treat BucketAlreadyOwnedByYou as success
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Projects

None yet

Development

Successfully merging this pull request may close these issues.