Hi Binjie, Thanks for bringing up this discussion, I think leverage GitHub Action and Minicube/K3s is the first step we can do, but considering that GitHub Runner has 7G memory limitation(I don’t know about cpu limitation), I’m not sure if we can test Master HA mode on that.
And thanks Keyong for providing external K8s cluster, it’s really helpful for the quality of Celeborn. Thanks, Cheng Pan On Dec 7, 2022 at 17:44:04, wangsheng <[email protected]> wrote: > Hi Zhou, > > Thanks for your info about the External Kubernetes Cluster. > > I will enrich the solution in case of using External Kubernetes Cluster. > > From my opinion, using an external Kubernetes cluster will improve the > speed of CI by a large margin over using a built-in Kubernetes cluster. > > Anyway, I will enrich both solution, and bring these to Github issue. > > Thanks, > Binjie Yang > > 2022年12月7日 下午3:43，Keyong Zhou <[email protected]> 写道： > > > Hi sheng, > > > Thanks for the discussion. For the option of using External Kubernetes > > Cluster, currently we already have an external cluster for test purposes, > > maybe we can reuse the cluster. If necessary, we can also offer a small k8s > > cluster. > > > So if the external cluster is available, which solution do you prefer? > > > Thanks, > > Keyong Zhou > > > wangsheng <[email protected]> 于2022年12月6日周二 14:33写道： > > > > Hi Apache Celeborn(Incubating) Community Members: > > > > > > Current Apache Celeborn(Incubating), hereinafter referred to as Celeborn, > > > provides On Kubernetes capabilities and the corresponding Helm Chart, but > > > integration tests for On Kubernetes scenarios are missing. > > > > > > To ensure the availability and stability of the On Kubernetes case, we > > > need to add the above integration tests to the Github workflow. > > > > > > I have added some details in > > > > https://docs.google.com/document/d/1iIRFhXBFFsU6NX_zfi7kuCEYnh6NHUZh1KpOqSSp16s/edit?usp=sharing > > > < > > > > https://docs.google.com/document/d/1iIRFhXBFFsU6NX_zfi7kuCEYnh6NHUZh1KpOqSSp16s/edit?usp=sharing > >, > > > welcome your discussion and comments. > > > > > > Best, > > > Binjie Yang > > > >
