Virtualisation, Storage and various other ramblings.

Category: Kubernetes (Page 1 of 2)

Creating Kubernetes Clusters with Rancher and Pulumi

tldr; Here is the code repo

Intro

My Job at Suse (via Rancher) involves hosting a lot of demos, product walk-throughs and various other activities that necessitate spinning up tailored environments on-demand. To facilitate this, I previously leaned towards Terraform, and have since curated a list of individual scripts I have to manage on an individual basis as they address a specific use case.

This approach reached a point where it became difficult to manage. Ideally, I wanted an IaC environment that catered for:

  • Easy, in-code looping (ie for and range)
  • “Proper” condition handling, ie if monitoring == true, install monitoring vs the slightly awkward HCL equivalent of repurposing count as a sudo-replacement for condition handling.
  • Influence what’s installed by config options/vars.
  • Complete end-to end creation of cluster objects, in my example, create:
    • AWS EC2 VPC
    • AWS Subnets
    • AWS AZ’s
    • AWS IGW
    • AWS Security Group
    • 1x Rancher provisioned EC2 cluster
    • 3x single node K3S clusters used for Fleet
Architectural Overview

Pulumi addresses these requirements pretty comprehensively. Additionally, I can re-use existing logic from my Terraform code as the Rancher2 Pulumi provider is based on the Terraform implementation, but I can leverage Go tools/features to build my environment.

Code Tour – Core

The core objects are created directly, using types from the Pulumi packages:

VPC:

// Create AWS VPC
vpc, err := ec2.NewVpc(ctx, "david-pulumi-vpc", &ec2.VpcArgs{
	CidrBlock:          pulumi.String("10.0.0.0/16"),
	Tags:               pulumi.StringMap{"Name": pulumi.String("david-pulumi-vpc")},
	EnableDnsHostnames: pulumi.Bool(true),
	EnableDnsSupport:   pulumi.Bool(true),
})

You will notice some interesting types in the above – such as pulumi.Bool and pulumi.String. The reason for this is, we need to treat cloud deployments as asynchronous operations. Some values we will know at runtime (expose port 80), some will only be known at runtime (the ID of a VPC, as below). These Pulumi types are a facilitator of this asynchronous paradigm.

IGW

// Create IGW
igw, err := ec2.NewInternetGateway(ctx, "david-pulumi-gw", &ec2.InternetGatewayArgs{
	VpcId: vpc.ID(),
})

Moving to something slightly more complex, such as looping around regions and assigning a subnet to each:

// Get the list of AZ's for the defined region
azState := "available"
zoneList, err := aws.GetAvailabilityZones(ctx, &aws.GetAvailabilityZonesArgs{
	State: &azState,
})

if err != nil {
	return err
}

//How many AZ's to spread nodes across. Default to 3.
zoneNumber := 3
zones := []string{"a", "b", "c"}

var subnets []*ec2.Subnet

// Iterate through the AZ's for the VPC and create a subnet in each
for i := 0; i < zoneNumber; i++ {
	subnet, err := ec2.NewSubnet(ctx, "david-pulumi-subnet-"+strconv.Itoa(i), &ec2.SubnetArgs{
		AvailabilityZone:    pulumi.String(zoneList.Names[i]),
		Tags:                pulumi.StringMap{"Name": pulumi.String("david-pulumi-subnet-" + strconv.Itoa(i))},
		VpcId:               vpc.ID(),
		CidrBlock:           pulumi.String("10.0." + strconv.Itoa(i) + ".0/24"),
		MapPublicIpOnLaunch: pulumi.Bool(true),
	})

This is repeated for each type

Code Tour – Config

The config file allows us to store information required by providers (unless using env variables or something externally) and values that we can use to influence the resources that are created. In particular, I added the following boolean values:

config:
  Rancher-Demo-Env:installCIS: false
  Rancher-Demo-Env:installIstio: false
  Rancher-Demo-Env:installLogging: false
  Rancher-Demo-Env:installLonghorn: false
  Rancher-Demo-Env:installMonitoring: false
  Rancher-Demo-Env:installOPA: false
  Rancher-Demo-Env:installFleetClusters: false

This directly influence what will be created in my main demo cluster, as well as individual “Fleet” clusters. Within the main Pulumi code, these values are extracted:

conf := config.New(ctx, "")
InstallIstio := conf.GetBool("installIstio")
installOPA := conf.GetBool("installOPA")
installCIS := conf.GetBool("installCIS")
installLogging := conf.GetBool("installLogging")
installLonghorn := conf.GetBool("installLonghorn")
installMonitoring := conf.GetBool("installMonitoring")
installFleetClusters := conf.GetBool("installFleetClusters")

Because of this, native condition handling can be leveraged to influence what’s created:

if installIstio {
	_, err := rancher2.NewAppV2(ctx, "istio", &rancher2.AppV2Args{
		ChartName:    pulumi.String("rancher-istio"),
		ClusterId:    cluster.ID(),
		Namespace:    pulumi.String("istio-system"),
		RepoName:     pulumi.String("rancher-charts"),
		ChartVersion: pulumi.String("1.8.300"),
	}, pulumi.DependsOn([]pulumi.Resource{clusterSync}))

	if err != nil {
		return err
	}
}

As there’s a much more dynamic nature to this project, I have a single template which I can tailor to address a number of use-cases with a huge amount of customisation. One could argue the same could be done in Terraform with using count, but I find this method cleaner. In addition, my next step is to implement some testing using go’s native features to further enhance this project.

Bootstrapping K3s

One challenge I encountered was being able to create and import K3s clusters. Currently, only RKE clusters can be directly created from Rancher. To address this, I created the cluster object in Rancher, extract the join command, and passed it together with the K3s install script so after K3s has stood up, it will run the join command:

if installFleetClusters {
	// create some EC2 instances to install K3s on:
	for i := 0; i < 3; i++ {
		cluster, _ := rancher2.NewCluster(ctx, "david-pulumi-fleet-"+strconv.Itoa(i), &rancher2.ClusterArgs{
			Name: pulumi.String("david-pulumi-fleet-" + strconv.Itoa(i)),
		})

		joincommand := cluster.ClusterRegistrationToken.Command().ApplyString(func(command *string) string {
			getPublicIP := "IP=$(curl -H \"X-aws-ec2-metadata-token: $TOKEN\" -v http://169.254.169.254/latest/meta-data/public-ipv4)"
			installK3s := "curl -sfL https://get.k3s.io | INSTALL_K3S_VERSION=v1.19.5+k3s2 INSTALL_K3S_EXEC=\"--node-external-ip $IP\" sh -"
			nodecommand := fmt.Sprintf("#!/bin/bash\n%s\n%s\n%s", getPublicIP, installK3s, *command)
			return nodecommand
		})

		_, err = ec2.NewInstance(ctx, "david-pulumi-fleet-node-"+strconv.Itoa(i), &ec2.InstanceArgs{
			Ami:                 pulumi.String("ami-0ff4c8fb495a5a50d"),
			InstanceType:        pulumi.String("t2.medium"),
			KeyName:             pulumi.String("davidh-keypair"),
			VpcSecurityGroupIds: pulumi.StringArray{sg.ID()},
			UserData:            joincommand,
			SubnetId:            subnets[i].ID(),
		})

		if err != nil {
			return err
		}
	}

}

End result:

     Type                               Name                                  Status       
 +   pulumi:pulumi:Stack                Rancher-Demo-Env-dev                  creating...  
 +   pulumi:pulumi:Stack                Rancher-Demo-Env-dev                  creating..   
 +   pulumi:pulumi:Stack                Rancher-Demo-Env-dev                  creating..   
 +   ├─ rancher2:index:Cluster          david-pulumi-fleet-1                  created      
 +   ├─ rancher2:index:Cluster          david-pulumi-fleet-2                  created      
 +   ├─ rancher2:index:CloudCredential  david-pulumi-cloudcredential          created      
 +   ├─ aws:ec2:Subnet                  david-pulumi-subnet-1                 created      
 +   ├─ aws:ec2:Subnet                  david-pulumi-subnet-0                 created      
 +   ├─ aws:ec2:InternetGateway         david-pulumi-gw                       created     
 +   ├─ aws:ec2:Subnet                  david-pulumi-subnet-2                 created     
 +   ├─ aws:ec2:SecurityGroup           david-pulumi-sg                       created     
 +   ├─ aws:ec2:DefaultRouteTable       david-pulumi-routetable               created     
 +   ├─ rancher2:index:NodeTemplate     david-pulumi-nodetemplate-eu-west-2b  created     
 +   ├─ rancher2:index:NodeTemplate     david-pulumi-nodetemplate-eu-west-2a  created     
 +   ├─ rancher2:index:NodeTemplate     david-pulumi-nodetemplate-eu-west-2c  created     
 +   ├─ aws:ec2:Instance                david-pulumi-fleet-node-0             created     
 +   ├─ aws:ec2:Instance                david-pulumi-fleet-node-2             created     
 +   ├─ aws:ec2:Instance                david-pulumi-fleet-node-1             created     
 +   ├─ rancher2:index:Cluster          david-pulumi-cluster                  created     
 +   ├─ rancher2:index:NodePool         david-pulumi-nodepool-2               created     
 +   ├─ rancher2:index:NodePool         david-pulumi-nodepool-1               created     
 +   ├─ rancher2:index:NodePool         david-pulumi-nodepool-0               created     
 +   ├─ rancher2:index:ClusterSync      david-clustersync                     created     
 +   ├─ rancher2:index:AppV2            opa                                   created     
 +   ├─ rancher2:index:AppV2            monitoring                            created     
 +   ├─ rancher2:index:AppV2            istio                                 created     
 +   ├─ rancher2:index:AppV2            cis                                   created     
 +   ├─ rancher2:index:AppV2            logging                               created     
 +   └─ rancher2:index:AppV2            longhorn                              created     
 
Resources:
    + 29 created

Duration: 19m18s

20mins for a to create all of these resources fully automated is pretty handy. This example also includes all the addons – opa, monitoring, istio, cis, logging and longhorn.

K3s, Rancher and Pulumi

TLDR; Repo can be found here (Be warned, I’m at best, a hobbyist programmer and certainly not a software engineer in my day job)

I’ve been recently getting acquainted with Pulumi as an alternative to Terraform for managing my infrastructure. I decided to create a repo that would do a number of activities to stand up Rancher in a new K3s cluster, all managed by Pulumi in my vSphere Homelab, consisting of the following activities:

  • Provision three nodes from a VM Template.
  • Use cloud-init as a bootstrapping utility:
    • Install K3s on the first node, elected to initialise the cluster.
    • Leverage K3s’s Auto-Deploying Manifests feature to install Cert-Manager, Rancher and Metallb.
  • Join two additional nodes to the cluster to form a HA, embedded etcd cluster.

The Ingress Controller is exposed via a loadbalancer service type, leveraging Metallb.

After completion, Pulumi will output the IP address (needed to create a DNS record) and the specified URL:

tputs:
    Rancher IP (Set DNS): "172.16.10.167"
    Rancher url:        : "rancher.virtualthoughts.co.uk"

Why cloud-init ?

For this example, I wanted to have a zero-touch deployment model relative to the VM’s themselves – IE no SSH’ing directly to the nodes to remotely execute commands. cloud-init addresses these requirements by having a way to seed an instance with configuration data. This Pulumi script leverages this in two ways:

  1. To set the instance (and therefore host) name as part of metadata.yaml (which is subject to string replacement)
  2. To execute a command on boot that initialises the K3s cluster (Or join an existing cluster for subsequent nodes) as part of userdata.yaml
  3. To install cert-manager, rancher and metallb, also as part of userdata.yaml

Reflecting on Using Pulumi

Some of my observations thus far:

  • I really, really like having “proper” condition handling and looping. I never really liked repurposing count in Terraform as awkward condition handling.
  • Being able to leverage standard libraries from your everyday programming language makes it hugely flexible. An example of this was taking the cloud-init user and metadata and encoding it in base64 by using the encoding/base64 package.

End to end automation with CircleCI and ArgoCD Part 3 – ArgoCD

ArgoCD is a Continuous Delivery tool designed for Kubernetes, This will be used to take the generated YAML file from the CI process and apply it to two clusters.

In this section, the following part of the overall CI/CD pipeline being implemented is depicted below.

  • ArgoCD will monitor for changes in the Webapp-CD Github Repo.
  • All changes are automatically applied to the Test cluster.
  • All changes will be staged for manual approval to Prod cluster

Install ArgoCD

ArgoCD has extensive installation documentation here. For ease, a community Helm chart has also been created.

Add Clusters to ArgoCD

I’m using Rancher deployed clusters which require a bit of tweaking on the ArgoCD side. The following Gist outlines this well: https://gist.github.com/janeczku/b16154194f7f03f772645303af8e9f80. However, other clusters can be added with a argocd cluster add, which will leverage the current kubeconfig file. For this environment, both Prod and Test clusters were added.

If done correctly, the clusters should be visible in Settings > Clusters in the ArgoCD web UI and argocd cluster list in the CLI:

Add Repo to ArgoCD

In ArgoCD navigate to Settings > Repositories:

You can connect to a repo in one of two ways – SSH or HTTPS. Fill in the name URL and respective authentication parameters:

With all being well, it will be displayed:

Create ArgoCD Application

From the ArgoCD UI, select either New App or Create Application

Application Settings – General

Application Name: web-test
Project: Default (For production environments a separate project would be created with specific resource requirements, but this will suffice to get started with.)
Sync Policy: Automatic

Application Settings – Source

Repository URL: Select from dropdown
Revision: Head
Path: .

Application Settings – Destination

Cluster: Test
Namespace: Default (This is used when the application does explicitly define a namespace to reside in).

After which our application has been added:

Selecting the app will display its constituent parts

Repeat the create application process but substitute the following values:

Application Name : web-prod
Sync Policy: Manual
Cluster: Prod

After which, both the prod and test applications will be shown

webapp-prod is noted as being OutOfSync – this is expected. In this environment, I don’t want changes to automatically propagate to prod, but only to test. Clicking “Sync” will rectify this:

Testing

Now everything is in place the following should occur during a new commit to the source code:

  • Automatically invoke CircleCI pipeline to Test, Build and Publish the application as a container to Dockerhub
  • Construct a YAML file using the aforementioned image
  • ArgoCD detects a change in the manifest and:
    • Applies it to Test immediately
    • Reports Prod is now out of sync.

As expected, changes to the source code have propagated into the Kubernetes clusters it is residing on.

Part 1 – Overview

Part 2 – CircleCI Configuration

« Older posts

© 2021 Virtual Thoughts

Theme by Anders NorenUp ↑

Social media & sharing icons powered by UltimatelySocial
RSS
Twitter
Visit Us
Follow Me