noderesources: exempt annotated pods from max-pod-count cap#142
Draft
eakman-datadog wants to merge 9 commits into
Draft
noderesources: exempt annotated pods from max-pod-count cap#142eakman-datadog wants to merge 9 commits into
eakman-datadog wants to merge 9 commits into
Conversation
Datadog: **NOT FROM UPSTREAM K8S**. From Lyft: lyft@266a18a This commit is the combination of multiple commits we've done to cherry pick and bring lyft sidecars across the ages. datadog:patch Origin: lyft@266a18a Drop-After: never (Datadog-maintained Lyft sidecar support)
datadog:patch Origin: #87 Drop-After: never (Datadog-specific flowcontrol fix)
This should allow build to have fips crypto enabled when we build with CGO_ENABLED=1 datadog:patch Origin: Datadog internal Drop-After: never (Datadog FIPS/boringcrypto build requirement)
…IgnoredDuringExecution Signed-off-by: xyz-li <hui0787411@163.com> datadog:patch Origin: kubernetes#129315 (closed, never merged) Drop-After: never (upstream PR closed without merge)
Origin: #98 Drop-After: never (Datadog-specific resourcequota behavior)
datadog:patch Upstream-PR: kubernetes#133654 Drop-After: v1.36.0 (expected merge in 1.36)
When a token's kid is absent from the cached JWKS, go-oidc v2 only refetches if the HTTP cache TTL has expired. EKS sets max-age=604800 (7 days), so new signing keys added by AWS within that window are never picked up, causing CCM and other workloads whose tokens are signed with the new key to fail with "failed to verify id token signature" until the apiserver is restarted. Fix: discard the cache-TTL guard in verify(). If a kid is not found in the local cache, always attempt a remote fetch. Subsequent tokens signed with cached (old) keys continue to verify without a network roundtrip; only an unknown kid triggers a refetch. datadog:patch Origin: coreos/go-oidc#259 Drop-After: kubernetes#134454
Built on top of the GBI Ubuntu base image: registry.ddbuild.io/images/base/gbi-ubuntu_2404:release@sha256:de9ff286f1a8bde796fc2b4afde9134079d74d90abe8b7b615212838e8fc01fd
Datadog: **NOT FROM UPSTREAM K8s**. Adds a pod-count exemption for pods carrying the annotation kubelet.datadoghq.com/exclude-from-max-pods="true" (literal value, case-sensitive, fail-closed). To be exempt the pod must also be host networked. Exempt pods bypass the v1.ResourcePods / Allocatable.Pods check in fitsRequest but are subject to all other scheduler and kubelet admission checks unchanged. The exemption is resolved once at PreFilter time via isExcludedFromMaxPodCount (pod_count_exemption.go) and flows through the shared AdmissionCheck seam, so both the scheduler filter path and the kubelet admission path honour it without a kubelet-specific branch. datadog:patch
|
cd0d3d0 to
f624c41
Compare
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Datadog: NOT FROM UPSTREAM K8s.
The purpose of this draft PR is to display testing evidence for a custom patch on 1.34 I'm working on.
I tested the PR using kind
Get the current number of pods on the kind node and fill it to the max with filler pods
Test that non-exempt pods are not scheduled
Test that exempt pods get scheduled