Take the Wheel, Don’t Reinvent It! - Deploying Apps With Helm in 5 Minutes - Christopher Hanson, RX-M Enterprises, LL
So you’ve deployed Kubernetes, now what? Time to deploy your applications! First things first, describe a Deployment to control your Pods--it is probably going to need a Secret or two and maybe a ConfigMap too! Now specify a Service to give it a VIP and introduce an Ingress to expose it! Saving some data? Propose a PVC! What about adding a Service Account or CRB? Lost in alliterations and acronyms? Helm is here to help!
This lighthearted session introduces Helm in an approachable way: revealing its architecture and nomenclature; introducing Charts, Chart components, and the official Charts repository for curated apps; as well as explaining Releases and Release upgrades. Information will be presented in a platform agnostic way to show attendees how to read, customize, and deploy an existing Chart to fit their needs, whether in the cloud, on a given platform or in their own datacenter.
click here for slide Helm_ckhanson_KubeConChina
From Enterprise Image Registry to Chart Repository - Tan Jiang, VMware
HARBOR: as the Kubernetes Registry.
In the On-Prem enterprise environment, we've been packaging content into docker images and hosting them using docker registry for some time, now helm chart has become the choice of application package format on Kubernetes, it's time to explore what's the proper way to host Helm Charts.
Based on our experience extending a private docker image registry (Project Harbor) to a Helm Chart registry. We'll share some of the key issues we see in Helm Chart management in an enterprise, such as RBAC, signing, vulnerability scanning, and how we provide a integrated user experience in managing/inspecting Helm Charts with docker image management. We'll also discuss the challenges we are seeing in this journey, and the possible solutions.
click here for slide From Enterprise Image Registry to Chart Repository
Kubernetes Explained the Easy Way, Using Computer Games... - Steven Trescinski, Gluo
How do you explain some of the basic Kubernetes constructs (i.e. nodes, pods, rolling updates, image tags, deployments,...) together with their concrete use cases in a fun and memorable way to people that have never heard of Kubernetes before?
The above was the challenge I was faced with when preparing my talk titled "Orchestrating the Container Chaos" for a group of Belgian students. So I started looking for something that most engineering students can relate to...
I ended up modding a first-person shooting computer game to interact with the Kubernetes API to visualise some of the main Kubernetes constructs. During this talk I will explain the tools/technologies used and hopefully inspire others to think outside of the box when preparing for technical talks.
click here for slide Kubernetes Explained the Easy Way, Using Computer Games...
Upgrading Stateful Sets in Constrained Environment - Kishore Yerrapragada, A10 Networks
The session will share the experiences we have had and lessons we learned in the process of upgrading micro service based platform with stateful sets in a resource constrained environment.
We have adopted Kubernetes as platform to host as well as distribute our management control plan that manages lots of high through put network software/hardware. We use the same platform to host SaaS as well as ship it on custom hardware to our on premise customers. While Kubernetes allows us to do rolling upgrades with out impacting the availability, we need to make sure our upgrades don't require extra resources as they run on fixed baremetal. These stateful sets include Cassandra, Elastic Search with out shared storage. The session will cover important hurdles in these environments and share some of the techniques we have used to over come those.
All About Kubernetes Certification Programs - Christopher Hanson, RX-M Enterprises, LLC
This lightning talk aims to raise awareness about the Kubernetes certification programs: CKA, CKAD, KCSP and CKCP. The discussion will introduce what they are, how they differ from one another, and the value they provide for participating individuals/vendors as well as companies looking to engage a vetted service provider. Attendees will be exposed to resources that will assist them in understanding the paths to certification or conformance as well as the educational and community resources available to test takers for exam preparation.
click here for slide All About Kubernetes Certification Programs
How to Simplify Data Scientist’s Day in Huge Company Using Kubernetes - Igor Khapov, IBM
In this presentation, speaker shares his experience on creating Data Science Experience Local based on Kubernetes and how it can accelerate data analysis process. The presenter describes different ways of using docker and Kubernetes for data scientist’s everyday job. IBM uses a cluster environment for internal and external data science projects. We will review all requirements from analytic team to cluster environment to develop, test and manage data science models based on Jupyter, R and Zeppelin notebooks
click here for slide How to Simplify Data Scientist’s Day in Huge Company Using Kubernetes
Kubernetes Love Machine Learning, Even on Private Cloud - Hui Luo, VMware
Kubernetes has established as a good platform for machine learning workloads by extending support of accelerators like GPU, all major public cloud provider are offering GPU enabled Kubernetes services, but public cloud is not the only option for users. There are ongoing efforts from the community to make running machine learning workloads with Kubernetes on private cloud as easy as on public cloud.
This talk is going to cover 3 major challenges that facing private cloud when enable GPU on Kubernetes. I will also demonstrate and discuss some of the projects that help to solve those challenges:
- Private cloud usually needs to support a wider range of GPU types, in some case, to support heterogeneous GPU in one cluster
- To support complex hardware topology like RDMA, NVLINK
- GPU resource contention is usually very high when limited GPU resource shared by multiple teams
click here for slideKubernetes Love Machine Learning, Even on Private Cloud
CI/CD Pipelines and Machine Learning - Jeremy Lewi, Google
This sessions will be part of the larger Kubeflow presentation. This will specifically cover how to use CI/CD pipelines to serve machine learning models on Kubernetes via Kubeflow.
Machine Learning requires a lot of training, experiments and retraining. It also requires a lot of data. Automating these can be difficult for many data scientists. Well now that we are able to make Machine Learning more cloud native via Kubeflow, allowing it to run on Kubernetes, we should now make use of the tooling that is available in a cloud native architecture. While most of my demo will involve WeaveWorks Flux, I will try to speak in general terms revolving around best practices in building a cloud native machine learning automation pipeline.
click here for slide CI/CD Pipelines and Machine Learning
Kubernetes CI/CD Practices at Meetup - Rui Chen, Meetup
The presentation would be like case-study style to introduce the Kubernetes practices @ Meetup around CI/CD.
The agenda of the presentation would be divided into:
- Current infrastructure status @ Meetup
- Configuration stack
- Deployment flow and process
- Challenges with managing things at scale
- Road ahead
click here for slide Kubernetes CI/CD Practices at Meetup
Jidproxy: ContainerCI On Kubernetes - Feng Wang, JD.COM
Jidproxy is a tool developed by the engineers of JD.COM. It implements the function of compiling packages and building images through Jenkins and Kubernetes. Its main features are as follows:
- Support jenkins master high availability
- Expand and Reduce jenkins matster and slave dynamically
- Compile from source code and build image with source code or code package
- Storage the compile and build task metadata information to database
click here for slide Jidproxy: ContainerCI On Kubernetes
Canary Deployment With Prometheus - Andrey Markelov, Infobip
I want to share my success story about using Prometheus for canary deployment. Currently I work with Java micro-services and every micro-service is usually deployed to the production environment at least once a week. To prevent fatal errors for high-loaded applications we deploy our application to one node with reduced traffic and after verifying we deploy to all others. This approach is called "canary deployment" and for us it is based on Prometheus metrics. We compare the metrics from the deployed node with historical metrics of other nodes and make a decision about version stability. I will show how we accomplish it and how it works. Also I would like to go deep into implementation details: HA Proxy, health checks, Jenkins. The real metrics examples will be shown.
click here for slideCanary Deployment With Prometheus
Effectively Extending Kubernetes Resources - Fan Zhang, VMware
In the real world, developing and maintaining Kubernetes CustomResourceDefinitions (CRD) and controllers are tedious, but add significant value for the Kubernetes cluster. Instead of building one from scratch, Kubebuilder aims to simplify API development by providing the cli tool and SDK, making it possible for the developer to focus on target purpose only, without having to care about how the Kubernetes resources work. In this talk, Fan Zhang will explore an effective approach to rapidly build and publish Kubernetes API by kubebuilder, and review how it is used in extending Kkubernetes resources in running cluster.
click here for slide Effectively Extending Kubernetes Resources
Kube-Liveboard: Make Your Cluster Visible - Xinkun Xu, JD.COM
Kube-liveboard is a visible tool which can provide more details of the cluster status by analyzing log data in real time and watching the resources from api. It can offer these:
- the number, the cost time, the source ip of the request group by namespaces, resources or request methed in apiserver.
- the cost time of each steps and the schedule result in kube-scheduler.
- the statistics of containers in different states at different times.
- the changes in the number of containers and nodes group by their states or namespaces.
- the cost time of each steps during the lifecyle of the pod. With it, the administrators can understand the trend of cluster changes and the bottleneck of the performance. Also, the administrators can take advantage of the data to send alarms. The topic will show why it is developed and how it works.
click here for slide Kube-Liveboard: Make Your Cluster Visible
Designing Kubernetes Clusters That Run on Embedded Devices - Daniele Polencic, learnk8s
Kubernetes is designed to run on data centres and in the cloud. After all, it is engineered to squeeze every last drop of efficiency from your servers. But does that mean that you can only use it on real servers? What about smaller computers such as routers? Can you create a Kubernetes cluster that is made of a collection of routers over a 5G network? This talk explores the challenges of designing such cluster topology and what you should look for when you create a cluster with networking, memory and CPU constraints.
click here for slideDesigning Kubernetes Clusters That Run on Embedded Devices