OpenShift Proof of Concept
Table of Content
- Platform Details
- Link Collection
- High Availability Architecture
- Bill of Materials
- Example Inventory
This repository aims to host information needed to install Red Hat OpenShift Container Platform. Focus is on a High Availability reference installation.
- Red Hat Enterprise Linux 7.5
- Red Hat OpenShift Container Platform 3.9
- Logical Architecture and Bill of Materials, page 5 & 6
- Host Preparation
- Advanced Installation
High Availability Architecture
Bill of Materials
The mentioned node sizes (not masters or loadbalancer) are the minimum prerequisites as of the documentation. Based on my experience it is a good rule of thumb to double the node size for a normal user experience, quadruple for a very good user experience.
Very important: The resources you apply to an OpenShift cluster should be fully reserved. There should not be any resource stealing from other VMs or memory, cpu, storage, network over-commitment. Otherwise the Kubernetes resource quotas will have no effect and application developers can not rely on the performance of the platform.
All hosts listed in the inventory file must be resolvable to the interface where OpenShift traffic is intended.
Additionally to the hosts in the inventory there must be defined three DNS aliases;
- The internal api endpoint DNS, eg. internal-ocp.example.com, resolving to the loadbalancer nodes LAN address, which forwards the traffic to the masters
- The external api endpoint DNS, eg. ocp.example.com, resolving to an external/internet facing loadbalancer, which forwards traffic also to the masters
- A wildcard DNS, eg *.apps.example.com, that will be forwarded to the OpenShift routers
Entries could also be added via the host preparation to /etc/hosts.
This PoC makes use of Container Native Storage (CNS), which is GlusterFS and Heketi deployed inside OpenShift. Please make sure there is an additional disk available with at least 200GB space on the first three nodes. See the glusterfs group in the inventory file. Check and change the device name if needed, eg.
OpenShift deploys software loadbalancers (HAProxy) on infra nodes to route traffic to applications, however a company enterprise grade loadbalancer cluster is assumed to forward traffic to these OpenShift software loadbalancers and API of the master nodes.
The OpenShift provided "lb" node is not meant to be used in production environments as it lacks high availability features.
The OpenShift provided "lb" node is only intended to serve API traffic to the "master" nodes, but can also be used to route traffic to the OpenShift hosted router aka software loadbalancer HAProxy.