Repository navigation
manifests: CPU requests and 5 s probes for the control-plane pods - #2557
Open
Davanum Srinivas (dims) wants to merge 3 commits into
Open
Davanum Srinivas (dims) wants to merge 3 commits into
Davanum Srinivas (dims) wants to merge 3 commits into
Conversation
Tim Hockin (thockin)
previously approved these changes
Oct 13, 2026
Collaborator
|
needs rebase |
Collaborator
Author
on it! |
Davanum Srinivas (dims)
force-pushed
the
control-plane-resources
branch
2 times, most recently
from
October 14, 2026 00:37
83ba11c to
785af3e
Compare
Bowei Du (bowei)
added
area/dev-infra
area/network
area/reliability
kind/bug
Davanum Srinivas (dims)
force-pushed
the
control-plane-resources
branch
3 times, most recently
from
October 18, 2026 21:05
60e3563 to
1d099d0
Compare
2 tasks done
Davanum Srinivas (dims)
force-pushed
the
control-plane-resources
branch
5 times, most recently
from
October 23, 2026 15:03
aa1e13d to
9d58cf7
Compare
ate-api-server, atenet-router, ate-controller and the dataplane sidecars ran BestEffort, and the atelet reserved 50m. On a node packed with sandboxes they got almost no CPU, their 1 s probes timed out and kubelet restarted them mid-run. Give each a CPU request, and every liveness and readiness probe a 5 s timeout and a failure threshold of 6, as the postgres probes already have.
… node The e2e kind cluster is one 4 vCPU node, and the requests summed to about 7 CPUs together with the kubeadm pods (ate-api-server 2x1, atenet-router 1.5, atelet 1.05, atenet-egress 0.7, ate-controller 0.5, postgres and prometheus 0.36, kube-system 0.95), so ate-api-server stayed Pending and the install timed out. Lower the requests to 250m for ate-api-server, atenet-router, both envoys and atelet, and to 100m for ate-controller, which puts the node at about 3.2 CPUs requested; the probe settings stay. go test on cmd/ate-setup, cmd/atecontroller and the atenet manifest tests passes, and the kind overlay renders.
…n a 4 vCPU kind node The requests exist for Burstable QoS and CPU shares under contention, so small values do the job and operators size up with overlays. The kind e2e lane deploys demo pools beside ate-system and the remaining room on a 4 vCPU node is about 1.5 CPU.
Davanum Srinivas (dims)
force-pushed
the
control-plane-resources
branch
from
October 24, 2026 11:56
9d58cf7 to
ba2f092
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.
ate-api-server, atenet-router, ate-controller, the dataplane sidecars and the atelet ran BestEffort (atelet: 50m) beside the sandboxes; under load their 1 s probes timed out and kubelet restarted them mid-run.
Give each a CPU request, and every liveness and readiness probe a 5 s timeout and a failure threshold of 6, as postgres already has. The requests are small, 100m (ate-controller 50m): they buy Burstable QoS and CPU shares under contention and leave a 4 vCPU kind node room for the demos; size up with an overlay.
Verified on a dense kind cluster: no control-plane restarts across a full run that previously had liveness kills.
go test ./cmd/ate-setup/... ./cmd/atecontroller/....