fix(docker): make kubectl-delivery base image and k8s version configurable#2
fix(docker): make kubectl-delivery base image and k8s version configurable#2vicoooo26 wants to merge 4 commits into
Conversation
|
🎉 Welcome to the Kubeflow Trainer! 🎉 Thanks for opening your first PR! We're happy to have you as part of our community 🚀 Here's what happens next:
Join the community:
Feel free to ask questions in the comments if you need any help or clarification! |
50b5d5b to
3b339d4
Compare
cheyang
left a comment
cheyang
left a comment
There was a problem hiding this comment.
Verified this locally: built the image with defaults (→ alpine 3.22.5, kubectl v1.30.7) and with --build-arg ALPINE_VERSION=3.20 --build-arg K8S_VERSION=v1.29.10 (→ alpine 3.20.10, kubectl v1.29.10). The global ARG ALPINE_VERSION declared before the first FROM is correctly referenced by both stages. Looks good.
Optional, non-blocking: the kubectl download still isn’t checksum-verified (pre-existing). Since you’re already touching this file, adding a sha256 check would be a nice hardening.
There was a problem hiding this comment.
Pull request overview
This PR makes the kubectl-delivery container image easier to customize by parameterizing previously hardcoded versions in the Dockerfile, while also upgrading the default Alpine base image.
Changes:
- Introduces
ARG ALPINE_VERSION=3.22and uses it for both build and runtime stages (FROM alpine:${ALPINE_VERSION}). - Switches Kubernetes version from a hardcoded
ENVtoARG K8S_VERSION=v1.30.7so it can be overridden at build time.
💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.
kubernetes 36.0.0 regenerated the sync client with Pydantic models, breaking the FakeResponse deserialization hack used in SDK unit tests. Pin kubernetes<36 in all CI workflows that install the Python SDK (test-python, integration-tests, e2e-test-train-api, test-example-notebooks) to avoid impacting SDK release. Add a TODO for future Pydantic work. Signed-off-by: Vico Chu <vico24826@gmail.com>
JAX worker pods were OOMKilled (exit 137) with 3Gi memory limit. Increase to 5Gi to accommodate XLA compilation and training memory usage. GitHub-hosted runners have 16GB RAM, so 2 replicas × 5Gi is feasible. Signed-off-by: Vico Chu <vico24826@gmail.com>
…rable Add ALPINE_VERSION (default 3.22) and K8S_VERSION (default v1.30.7) as build args so they can be overridden via --build-arg without editing the Dockerfile. Signed-off-by: Vico Chu <vico24826@gmail.com>
Download the official .sha256 from dl.k8s.io and verify the binary before installing, preventing use of a tampered kubectl.
6ccb670 to
f2ada4d
Compare
What this PR does / why we need it:
Makes the kubectl-delivery Dockerfile more flexible by extracting hardcoded Alpine and Kubernetes versions into build args:
ALPINE_VERSION(default3.22) — replaces the hardcodedalpine:3.17in both build and final stagesK8S_VERSION(defaultv1.30.7) — replaces the hardcodedENV K8S_VERSIONwith anARG, allowing override via--build-argThis also upgrades the Alpine base from 3.17 to 3.22, bringing security fixes and updated packages. Both values can now be overridden at build time without editing the Dockerfile.