Kubernetes sig-node (Asia) meeting notes

Kubernetes 1.2 Status Update (@dchen)

Huawei Conformance Test (@liangchenye)


Hyper integration (@feiskyer)

Hypernetes e2e test有一个单机环境的测试结果,但由于GCE环境的问题,多机环境的搭建还没有完成。还有一个Hyper日志”\n”处理的bug需要修复。


etcd计划在4月份发布v3.0 alpha,并争取kubernetes v1.3升级到etcd 3.0 stable。

More notes taken by @lantao

Release Brain Storm

@hongchao: Over-commit? -> In 1.3 roadmap today, but not finalized.In fact we can do over-commit today, but reliablility and QoS should be guranteed.(@dchen) Detect out of resource Decide which pod should be killed Well-defined eviction policy is needed.

Q&A @dchen: Etcd v3.0 alpha version in April; stable version in May or June; Try to ship with kubernetes 1.3.

@xiang: New node metric api merged? Yeah. Is the new metric api added for better scheduling? No, necessary information for scheduling is available on kubelet side from v1.0. Hardware level metrics?

@feiskyer: Pod level data in cadvisor? At cgroup level, we could do that. But at API level, it is still in disicussion. Internally shareable resource between containers is very powerful but quite complex. If customers don’t ask for that now, we may want to defer it. @dchen: How to define extra overhide of rkt and hyper in api, should it be visible for user? Related issue: https://github.com/coreos/rkt/issues/1788

Previous notes is recorded at https://docs.google.com/document/d/1L8s6Nyu5hNJxCOZLqJsuVEFAScWKhbwcis0X-j-upDc/edit#heading=h.jgbd25swf99j


comments powered by Disqus