Working with Red Hat OpenShift on vSphere, I’m really starting to understand the main infrastructure components and how everything fits together.
Next up was understanding how to control the cluster size after initial deployment. So, with Red Hat OpenShift, there are some basic concepts we need to understand first, before we jump into the technical how-to’s below in this blog.
In this blog I will cover the following;
- Understanding the concepts behind controlling Machines in OpenShift
- Editing your MachineSet to control your Virtual Machine Resources
- Editing your MachineSet to scale your cluster manually
- Deleting a node
- Configuring ClusterAutoscaler to automatically scale your environment
Machine API
The Machine API is a combination of primary resources that are based on the upstream Cluster API project and custom OpenShift Container Platform resources.
The Machine API performs all node host provisioning management actions as a post cluster installation method, providing you dynamic provisioning on top of your VMware vSphere platform (and other public/private cloud platforms).
The two primary resources are:
Machines
An object that describes the host for a Node. A machine has a providerSpec, which describes the types of compute nodes that are offered for different cloud platforms. For example, a machine type for a worker node on Amazon Web Services (AWS) might define a specific machine type and required metadata.
MachineSets
Groups of machines. MachineSets are to machines as ReplicaSets are to Pods. If you need more machines or must scale them down, you change the replicas field on the MachineSet to meet your compute need.
These custom resources add capabilities to your OpenShift cluster:
MachineAutoscaler
This resource automatically scales machines in a cloud. You can set the minimum and maximum scaling boundaries for nodes in a specified MachineSet, and the MachineAutoscaler maintains that range of nodes. The MachineAutoscaler object takes effect after a ClusterAutoscaler object exists. Both ClusterAutoscaler and MachineAutoscaler resources are made available by the ClusterAutoscalerOperator.
ClusterAutoscaler
This resource is based on the upstream ClusterAutoscaler project. In the OpenShift Container Platform implementation, this is integrated with the Machine API by extending the MachineSet API. You can set cluster-wide scaling limits for resources such as cores, nodes, memory, GPU, etc. You can configure priorities so that the cluster prioritizes pods so that new nodes are not brought online for less important pods. You can also set the ScalingPolicy, so that for example, you can scale up nodes but not scale down the node count.
MachineHealthCheck
This resource detects when a machine is unhealthy, deletes it, and, on supported platforms, creates a new machine. You can read more here about this technology preview feature in OCP 4.6.
Editing your MachineSet to control your Virtual Machine Resources
This blog post is an accompaniment to the session “vSphere with Tanzu Kubernetes – Day 2 Operations for the VI Admin” created by myself and Simon Conyard, with special thanks to the VMware LiveFire Team for allowing us access to their lab environments to create the technical demo recordings.
You can see the full video with technical demos below (1hr 4 minutes). This blog post acts a supplement to the recording.
This session is 44 minutes long (and is a little shorter than the one above).
The basic premise of the presentation was set at around a level-100/150 introduction to the Kubernetes world and marrying that to your knowledge of VMware vSphere as a VI Admin. Giving you an insight into most of the common areas you will need to think about when all of a sudden you are asked to deploy Tanzu Kubernetes and support a team of developers.
Scene Setting
So why are we talking about VMware and Kubernetes? Isn’t VMware the place where I run those legacy things called virtual machines?
Essentially the definition of an application has changed. On the left of the below image, we have the typical Application, we usually talk about the three tier model (Web, App, DB).
However, the landscape is moving towards the right hand side, applications running more like distributed systems. Where the data your need to function is being served, serviced, recorded, and presented not only by virtual machines, but Kubernetes services as well. Kubernetes introduces its own architectures and frameworks, and finally this new buzzword, serverless and functions.
Although you may not be seeing this change happen immediately in your workplace and infrastructure today. It is the direction of the industry.
Within vSphere there are two types of Kubernetes clusters that run natively within ESXi.
Supervisor Kubernetes cluster control plane for vSphere
Tanzu Kubernetes Cluster, sometimes also referred to as a “Guest Cluster.”
Supervisor Kubernetes Cluster
This is a special Kubernetes cluster that uses ESXi as its worker nodes instead of Linux.
This is achieved by integrating the Kubernetes worker agents, Spherelets, directly into the ESXi hypervisor. This cluster uses vSphere Pod Service to run container workloads natively on the vSphere host, taking advantage of the security, availability, and performance of the ESXi hypervisor.
To learn more about how vSphere delivers Kubernetes natively
The supervisor cluster is not a conformant Kubernetes cluster, by design, using Kubernetes to enhance vSphere. This ultimately provides you the ability to run pods directly on the ESXi host alongside virtual machines, and as the management of Tanzu Kubernetes Clusters.
Tanzu Kubernetes Cluster
To deliver Kubernetes clusters to your developers, that are standards aligned and fully conformant with upstream Kubernetes, you can use Tanzu Kubernetes Clusters (also referred to as “Guest” clusters.)
A Tanzu Kubernetes Cluster is a Kubernetes cluster that runs inside virtual machines on the Supervisor layer and not on vSphere Pods.
As a fully upstream-compliant Kubernetes it is guaranteed to work with all your Kubernetes applications and tools. Tanzu Kubernetes Clusters in vSphere use the open source Cluster API project for lifecycle management, which in turn uses the VM Operator to manage the VMs that make up the cluster.
Supervisor Cluster or Tanzu Kubernetes Cluster, which one should I choose to run my application?
Supervisor Cluster:
Has additional capabilities that are inherent in the vSphere environment and are available to Kubernetes via the kubectl command
Provides the ability to manage containers just as you would manage virtual machines
Provides stronger security and resource isolation due to the use of vSphere Pods
VMware Cloud Foundation is an integrated full stack solution, delivering customers a validated architecture bringing together vSphere, NSX for software defined networking, vSAN for software defined storage, and the vRealize Suite for Cloud Management automation and operation capabilities.
Deploying the vSphere Tanzu Kubernetes solution is as simple as a few clicks in a deployment wizard, providing you a fully integrated Kubernetes deployment into the VMware solutions.
Don’t have VCF? Then you can still enable Kubernetes yourself in your vSphere environment using vSphere 7.0 U1 and beyond. There will be extra steps for you to do this, and some of the integrations to the VMware software stack will not be automatic.
The below graphic summarises the deployment steps between both options discussed.
Building on top of the explanation of Tanzu Kubernetes Cluster explained earlier, Tanzu Kubernetes Grid (TKG) is the same easy-to-upgrade, conformant Kubernetes, with pre-integrated and validated components. This multi-cloud Kubernetes offering that you can run both on-premises in vSphere and in the public cloud on Amazon and Microsoft Azure, fully supported by VMware.
Tanzu Kubernetes Grid (TKG) is the name used for the deployment option which is multi-cloud focused.
Tanzu Kubernetes Cluster (TKC) is the name used for a Tanzu Kubernetes deployment deployed and managed by vSphere Namespace.
Introducing vSphere Namespaces
When enabling Kubernetes within a vSphere environment a supervisor cluster is created within the VMware Data Center. This supervisor cluster is responsible for managing all Kubernetes objects within the VMware Data Center, including vSphere Namespaces. The supervisor cluster communicating with ESXi forms the Kubernetes control plane, for enabled clusters.
A vSphere Namespace is a logical object that is created on the vSphere Kubernetes supervisor cluster. This object tracks and provides a mechanism to edit the assignment of resources (Compute, Memory, Storage & Network) and access control to Kubernetes resources, such as containers or virtual machines.
You can provide the URL of the Kubernetes control plane to developers as required, where they can then deploy containers to the vSphere Namespaces for which they have permissions.
Resources and permissions are defined on a vSphere Namespace for both Kubernetes containers, consuming resources directly via vSphere, or Virtual Machines configured and provisioned to operate Tanzu Kubernetes Grid (TKG).
For a Virtual Administrator the way access can be assigned to various Tanzu elements within the Virtual Infrastructure is very similar to any other logical object.
Create Roles
Assign Permissions to the Role
Allocate the Role to Groups or Individuals
Link the Group or Individual to inventory objects
With Tanzu those inventory objects include Namespaces’ and Resources.
What I also wanted to highlight was if a Virtual Administrator gave administrative permissions to a Kubernetes cluster, then this has similarities to granting ‘root’ or ‘administrator’ access to a virtual machine. An individual with these permissions could create and grant permissions themselves, outside of the virtual infrastructure.
Here I updated the vSphere CSI driver to work the additional security constraints that are baked into OpenShift 4.x.
Since then, once of the things that has been on my list to test is file volumes backed by vSAN File shares. This feature is available in vSphere 7.0.
Well I’m glad to report it does in fact work, by using my CSI driver (see above blog or my github), you can simply deploy consume VSAN File services, as per the documentation here.
I’ve updated my examples in my github repository to get this working.
OK just tell me what to do…
First and foremost, you need to add additional configuration to the csi conf file (csi-vsphere-for-ocp.conf).
If you do not, the defaults will be assumed which is full read-write access from any IP to the file shares created.
[Global]
# run the following on your OCP cluster to get the ID
# oc get clusterversion -o jsonpath='{.items[].spec.clusterID}{"\n"}'
cluster-id = c6d41ba1-3b67-4ae4-ab1e-3cd2e730e1f2
[NetPermissions "A"]
ips = "*"
permissions = "READ_WRITE"
rootsquash = false
[VirtualCenter "10.198.17.253"]
insecure-flag = "true"
user = "[email protected]"
password = "Admin!23"
port = "443"
datacenters = "vSAN-DC"
targetvSANFileShareDatastoreURLs = "ds:///vmfs/volumes/vsan:52c229eaf3afcda6-7c4116754aded2de/"
Next, create a storage class which is configured to consume VSAN File services.
Note2: December 2021 VMware released the Red Hat Certified Operator "vSphere Kubernetes Driver Operator", which is now the preferred and recommended way to install CPI and CSI in your OpenShift environment.
- Using the new vSphere Kubernetes Driver Operator with Red Hat OpenShift via Operator Hub
Note: This blog post was updated in February 2021 to use the new driver manifests from the Official VMware CSI Driver repository, which now provides support for OpenShift
Introduction
In this post I am going to install the vSphere CSI Driver version 2.1.0 with OpenShift 4.x, in my demo environment I’m connecting to a VMware Cloud on AWS SDDC and vCenter, however the steps are the same for an on-prem deployment.
We will be using the vSphere CSI Driver which now supports OpenShift.
- Pre-Reqs
- - vCenter Server Role
- - Download the deployment files
- - Create the vSphere CSI secret in OpenShift
- - Create Roles, ServiceAccount and ClusterRoleBinding for vSphere CSI Driver
- Installation
- - Install vSphere CSI driver
- - Verify Deployment
- Create a persistent volume claim
- Using Labels
- Troubleshooting
In your environment, cluster VMs will need “disk.enableUUID” and VM hardware version 15 or higher.
Pre-Reqs
vCenter Server Role
In my environment I will use the default administrator account, however in production environments I recommend you follow a strict RBAC procedure and configure the necessary roles and use a dedicated account for the CSI driver to connect to your vCenter.
To make life easier I have created a PowerCLI script to create the necessary roles in vCenter based on the vSphere CSI documentation;
Create the vSphere CSI Secret + CPI ConfigMap in OpenShift
Edit the two files “csi-vsphere.conf” + “vsphere.conf” with your vCenter infrastructure details. These two files may have the same information in them, but in the example of using VSAN File Services, then you may include further configuration in your CSI conf file, as an example.
[Global]
# run the following on your OCP cluster to get the ID
# oc get clusterversion -o jsonpath='{.items[].spec.clusterID}{"\n"}'
#Your OCP cluster name provided below can just be a human readable name but needs to be unique when running different OCP clusters on the same vSphere environment.
cluster-id = "OCP_CLUSTER_ID"
[VirtualCenter "VC_FQDN"]
insecure-flag = "true"
user = "USER"
password = "PASSWORD"
port = "443"
datacenters = "VC_DATACENTER"
Create the CSI secret + CPI configmap;
oc create secret generic vsphere-config-secret --from-file=csi-vsphere.conf --namespace=kube-system
oc create configmap cloud-config --from-file=vsphere.conf --namespace=kube-system
To validate:
oc get secret vsphere-config-secret --namespace=kube-system
oc get configmap cloud-config --namespace=kube-syste
This configuration is for block volumes, it is also supported to configure access to VSAN File volumes, and you can see an example of the configuration here;
You can verify the deployment with the two below commands
oc get deployments --namespace=kube-system
oc get CSINode
Creating a Storage Class that uses the CSI-Driver
Create a storage class to test the deployment. As I am using VMC as my test environment, I must use some additional optional parameters to ensure that I use the correct VSAN datastore (WorkloadDatastore). You can visit the references below for more information.
In the VMC vCenter UI, you can get this by going to the Datastore summary page.
To get my datastore URL I need to reference, I will use PowerCLI
get-datastore work* | Select -ExpandProperty ExtensionData | select -ExpandProperty Info
I’m going to create my StorageClass on the fly, but you can find my example YAMLs here;
You can see the PVC created under my cluster > Monitor Tab > Cloud Native Storage in vCenter.
Using Labels
Thanks to one of my colleagues (Jason Monger), who asked me if we could use labels with this integration. And the answer is yes you can.
When creating your PVC, under metadata including your labels such as the able below. These will be pulled into your vCenter UI making it easier to associate your volumes.
For troubleshooting, you need to be aware of the four main containers that run in the vSphere CSI Controller pod and you should investigate the logs from these when you run into issues;
CSI-Attacher
CSI-Provisoner
vSphere-CSI-Controller
vSphere-Syncer
Below I have uploaded some of the logs from a successful setup and creation of a persistent volume.
The following procedure is intended to create VM’s from an OVA template booting with static IP’s when the DHCP server can not reserve the IP addresses.
The Problem
OCP requires that all DNS configurations be in place. VMware requires that the DHCP assign the correct IPs to the VM. Since many real installations require the coordination with different teams in an organization, many times we don’t have control of DNS, DHCP or Load balancer configurations.
The CoreOS documentation explain how to create configurations using ignition files. I created a python script to put the network configuration using the ignition files created by the openshift-install program.
Reference Architecture
For this guide, we are going to deploy 3 master nodes (control-plane) and 2 worker nodes (compute This guide uses RHEL CoreOS 4.3 as the virtual machine image, deploying Red Hat OCP 4.3, as per the support of N-1 from Red Hat.
We will use a centralised Linux server (Ubuntu) that will perform the following functions;
The deployment will be semi-automated using Terraform, so that we can easily build configuration files used by the CoreOS VM’s that have Static IP settings.
Using a later version of Terraform will cause failures.
In the below screenshot, the script has created the “demo” domain folder and entered my records. It is important that you have PTR records setup for everything apart from the “etcd-X” records.
To provide the best experiences, we use technologies like cookies to store and/or access device information. Consenting to these technologies will allow us to process data such as browsing behaviour or unique IDs on this site. Not consenting or withdrawing consent, may adversely affect certain features and functions.
Functional
Always active
The technical storage or access is strictly necessary for the legitimate purpose of enabling the use of a specific service explicitly requested by the subscriber or user, or for the sole purpose of carrying out the transmission of a communication over an electronic communications network.
Preferences
The technical storage or access is necessary for the legitimate purpose of storing preferences that are not requested by the subscriber or user.
Statistics
The technical storage or access that is used exclusively for statistical purposes.The technical storage or access that is used exclusively for anonymous statistical purposes. Without a subpoena, voluntary compliance on the part of your Internet Service Provider, or additional records from a third party, information stored or retrieved for this purpose alone cannot usually be used to identify you.
Marketing
The technical storage or access is required to create user profiles to send advertising, or to track the user on a website or across several websites for similar marketing purposes.
