Wei Huang

Software Engineer  | Co-chair of Kubernetes SIG-Scheduling | exIBMer | Opinions my own

[paco xu @xu_paco]

I'm at A Comparative Analysis of Kueue, Volcano, and YuniKorn - Wei Huang, Apple & Shiming Zhang, DaoCloud @sched # https://t.co/w9yBaOqF4R Welcome to Room G. It will start in 20mins.

[ahmet alp balkan @ahmetb]

I wrote a blog about some pitfalls we've learned generating Kubernetes Custom Resources with `controller-gen` over the years, and how to do more rigid validation and defaulting with it.
https://t.co/Y8g9p6LUd3

[Wei Zhang @kweizh]

Scheduler Plugins can now easily be installed by using Helm directly🚀 @hweicdl https://t.co/zXnC1FhG51

[Wei Huang @hweicdl]

Happy #kuberTENes! Check out my #FirstContribution to Kubernetes #ContribCard https://t.co/Vtd5hPUbEg

[Wei Huang @hweicdl]

Unusual Friday: spot a scheduler regression in 1.27 and provided a fix, brainstormed with @sanpo_shiho on another general issue, and surprisingly found another regression in 1.26...

[William Lam @lamw]

KWOK (Kubernetes WithOut Kubelet) project is freaking DOPE! https://t.co/HdXF99F0mV

Quickly mock up k8s cluster/nodes/pods literally in seconds!

Just deployed 50 nodes w/quick for loop 🔥

[Container Registry @ContainerRgstry]

Give KWOK (Kubernetes WithOut Kubelet) a try to quickly test things on Kubernetes local or in your CI/CD. Spin up a cluster in seconds as all Nodes and Pods are simulated. The approach employs a pretty low resource footprint and hence time.

https://t.co/7lXVduqKed

[K8sContributors @K8sContributors]

Blog: Kubernetes 1.26: Pod Scheduling Readiness - https://t.co/upaYx4fMGu #Kubernetes

[Wei Huang @hweicdl]

Shoutout to @AHGharaibeh @smarterclayton @liggitt for their dedicated review efforts! And @kerthcet @Itsuugo for the constructive comments!

[Wei Huang @hweicdl]

Just get the dev work done before the k8s 1.26 code freeze. Phew...

[Wei Huang @hweicdl]

The Silver Searcher (i.e., `ag`) is a must-have in my daily dev workflow. TIL: ag --<file-postfix> <string> (examples: ag --yaml <string>, ag --go <string>)

[Wei Huang @hweicdl]

TIL: child context spawned by context.WithCancel() are not GC-able. Check your code to ensure cancel() is always called. https://t.co/PiYvadm9dI

[Wei Huang @hweicdl]

TIL: https://t.co/fxu1urz2Ba, and how it manipulates API objects using a neat type `metav1.PartialObjectMetadata`.

[Wei Huang @hweicdl]

TIL: when manipulating CR objects that enabled subresource. Use controller-runtime's client.Update() to update the CR obj's spec, and use client.Status().Update() for the CR obj's status. Trying to update a CR obj's status via client.Update() would disappoint you.

[Wei Huang @hweicdl]

Should have promoted it more…

[ahmet alp balkan @ahmetb]

TIL Kubernetes has a framework to author custom schedulers that override only certain parts of the default scheduler. I somehow missed this. https://t.co/N0zkeScJ1p

[Wei Huang @hweicdl]

@ParisInBmore Niche Zero

[Wei Huang @hweicdl]

@ibuildthecloud @GoLandIDE Met exactly the same issue. The workaround works for me is to invalidate the local index cache…

[Yuan Chen @YuanChenByte]

I’ll attend #KubeCon in person next week. Here’s our talk on elastic quota support in #Kubernetes. The talk is scheduled at 11:55pm PDT on Wednesday, Oct. 13. Hope to see you. @sched # https://t.co/iiqk71bXEl

[Wei Huang @hweicdl]

setting the HTTP2_READ_IDLE_TIMEOUT_SECONDS env variable to 0: https://t.co/fgdr3PUXkI

[Wei Huang @hweicdl]

Tried debugging kube apiserver on my mac today. Everything went smooth except for a weird error from kubectl: "http2: client connection lost". Tried almost all timeout options but still didn't work... Until I found the so-called "http2 healthcheck" thing, which can be disabled by