Configuring Flux CD for GitOps

This document explains how you can configure Flux CD with Helm and GitHub integration to deploy a TSB application to a target cluster.

note

This document assumes that:

• Flux version 2 CLI is installed.
• Helm CLI is installed.
• TSB is up and running, and GitOps has been enabled for the target cluster.

Cluster setup

First, install Flux on the target cluster with a GitHub integration. To do that, use the following command under the target cluster kubernetes context.

note

You will need a GitHub Personal Access Token (PAT) to input in the following command.

$ flux bootstrap github \
  --owner=your-org \
  --repository=git-ops \
  --path=./clusters/cluster-01

note

Add --personal --private flags if you use your personal GitHub account for testing purposes.

This sets up Flux with the needed configurations for a cluster called cluster-01 in a GitHub repository named git-ops under the clusters/cluster-01/ directory.

note

Run flux logs -A --follow command in a different shell for debugging purposes.

You can run this command to do a generic status check:

$ flux check
► checking prerequisites
✔ Kubernetes 1.20.15-gke.2500 >=1.20.6-0
► checking controllers
✔ helm-controller: deployment ready
► ghcr.io/fluxcd/helm-controller:v0.20.1
✔ kustomize-controller: deployment ready
► ghcr.io/fluxcd/kustomize-controller:v0.24.2
✔ notification-controller: deployment ready
► ghcr.io/fluxcd/notification-controller:v0.23.4
✔ source-controller: deployment ready
► ghcr.io/fluxcd/source-controller:v0.24.0
✔ all checks passed

Under the hood, Flux uses a Kustomization with a GitRepository source to store its own resources.

You can query its status:

$ flux get all
NAME                        REVISION        SUSPENDED   READY   MESSAGE
gitrepository/flux-system   main/36dff73    False       True    stored artifact for revision 'main/36dff739b5ae411a7b4a64010d42937bd3ae4d25'

NAME                        REVISION        SUSPENDED   READY   MESSAGE
kustomization/flux-system   main/36dff73    False       True    Applied revision: main/36dff73

Meanwhile, you will see something like this in the logs:

2022-04-24T20:42:06.921Z info Kustomization/flux-system.flux-system - server-side apply completed
2022-04-24T22:51:30.431Z info GitRepository/flux-system.flux-system - artifact up-to-date with remote revision: 'main/36dff739b5ae411a7b4a64010d42937bd3ae4d25'

Since Flux is up and running now, the next step is to push new configurations to the git-ops repository for the cluster-01 cluster. You can clone the repository and cd into clusters/cluster-01 for the next steps.

Application setup

There are several ways to structure your GitOps repositories. In this example, and for simplicity reasons, the same repository is used for both cluster and application configurations.

The goal for this section is to deploy the Helm chart of a Bookinfo application and its TSB resources.

First, create the bookinfo namespace with sidecar injection:

kubectl create namespace bookinfo
kubectl label namespace bookinfo istio-injection=enabled

Then, create a HelmRelease Flux resource with a GitRepository source for Bookinfo.

note

The alternative to GitRepository is a HelmRepository, which is not covered in this document.

If the bookinfo TSB helm chart definition is stored in the apps/bookinfo directory, create the HelmRelease resource in clusters/cluster-01/bookinfo.yaml.

apiVersion: helm.toolkit.fluxcd.io/v2beta1
kind: HelmRelease
metadata:
  name: bookinfo
  namespace: flux-system
spec:
  chart:
    spec:
      chart: ./apps/bookinfo
      sourceRef:
        kind: GitRepository
        name: flux-system
  interval: 1m0s
  install:
    createNamespace: true
  targetNamespace: bookinfo

Note that:

• The HelmRelease will be created in the flux-system namespace, while the resources defined by the Helm chart of the apps/bookinfo chart of the release will be deployed in the bookinfo target namespace.
• Since spec.chart.spec.version is not specified, Flux will use latest chart version.
• GitRepository.name is flux-system since that's the name Flux uses internally for bootstrapping.

Next, add and push the file into git and watch the flux logs. You will see something like this:

2022-04-25T08:02:37.233Z info HelmRelease/bookinfo.flux-system - reconcilation finished in 49.382555ms, next run in 1m0s
2022-04-25T08:02:37.980Z info HelmChart/flux-system-bookinfo.flux-system - Discarding event, no alerts found for the involved object
2022-04-25T08:02:45.784Z error HelmChart/flux-system-bookinfo.flux-system - reconciliation stalled invalid chart reference: stat /tmp/helmchart-flux-system-flux-system-bookinfo-4167124062/source/apps/bookinfo: no such file or directory

This is because the helm chart has not been pushed to the apps/bookinfo directory yet.

Note that instead of parsing the flux logs, you can also query the resources with kubectl:

• kubectl get helmreleases -A
• kubectl get helmcharts -A

Next, create the helm chart. Create the apps/ directory, enter it and run:

$ helm create bookinfo

This creates the following file tree:

$ tree
.
+-- bookinfo
    +-- Chart.yaml
    +-- charts
    +-- templates
    |   +-- NOTES.txt
    |   +-- _helpers.tpl
    |   +-- deployment.yaml
    |   +-- hpa.yaml
    |   +-- ingress.yaml
    |   +-- service.yaml
    |   +-- serviceaccount.yaml
    |   \-- tests
    |       \-- test-connection.yaml
    \-- values.yaml

4 directories, 10 files

Then, cd into bookinfo/.

For simplicity reasons, remove the following not-needed content:

$ rm -rf values.yaml charts templates/NOTES.txt templates/*.yaml templates/tests/

Next, edit the Chart.yaml. A minimal content looks like this:

apiVersion: v2
name: bookinfo
description: TSB bookinfo Helm Chart.
type: application
version: 0.1.0
appVersion: "0.1.0"

Next, add the Bookinfo definitions to the templates/ directory, gathering them from Istio's repository:

curl https://raw.githubusercontent.com/istio/istio/master/samples/bookinfo/platform/kube/bookinfo.yaml -o bookinfo.yaml

Once we have the bookinfo deployment, we'll add the TSB configuration resources in a templates/tsb.yaml file. When creating the TSB configurations, a best practice is to put all them inside a List resource. This will enforce a strict order when applying them into the cluster, and you will be able to guarantee that the resources that are at higher levels of the TSB resource hierarchy are applied first, and you won't hit issues due to Helm resource ordering limitations.

note

For this example, an ingress gateway it's used for the application that will be configured by the first resource configuration below. You can read more about this here.

Also make sure you change your-org and your-tenant for actual values.

apiVersion: v1
kind: List
items:
# Create an ingress gateway deployment that will be the entry point to
# the bookinfo application
- apiVersion: install.tetrate.io/v1alpha1
  kind: IngressGateway
  metadata:
    namespace: bookinfo
    name: tsb-gateway-bookinfo
  spec: {}
# Create the workspace and gateway group that capture the namespaces where
# the bookinfo application will run
- apiVersion: tsb.tetrate.io/v2
  kind: Workspace
  metadata:
    name: bookinfo
    annotations:
      tsb.tetrate.io/organization: your-org
      tsb.tetrate.io/tenant: your-tenant
  spec:
    namespaceSelector:
      names:
        - "*/bookinfo"
- apiVersion: gateway.tsb.tetrate.io/v2
  kind: Group
  metadata:
    name: bookinfo-gg
    annotations:
      tsb.tetrate.io/organization: your-org
      tsb.tetrate.io/tenant: your-tenant
      tsb.tetrate.io/workspace: bookinfo
  spec:
    namespaceSelector:
      names:
        - "*/*"
    configMode: BRIDGED
# Expose the productpage service in the application ingress
- apiVersion: gateway.tsb.tetrate.io/v2
  kind: IngressGateway
  metadata:
    name: bookinfo-gateway
    annotations:
      tsb.tetrate.io/organization: your-org
      tsb.tetrate.io/tenant: your-tenant
      tsb.tetrate.io/workspace: bookinfo
      tsb.tetrate.io/gatewayGroup: bookinfo-gg
  spec:
    workloadSelector:
      namespace: bookinfo
      labels:
        app: tsb-gateway-bookinfo
    http:
      - name: productpage
        port: 80
        hostname: "bookinfo.example.com"
        routing:
          rules:
            - route:
                host: "bookinfo/productpage.bookinfo.svc.cluster.local"
                port: 9080
---

Before pushing, test that the chart is well constructed:

$ helm install bookinfo --dry-run .

It should print the rendered resources as YAML.

Now it's time to push them and check the flux logs.

If GitOps is properly configured in the cluster, pushing this chart will create the corresponding Kubernetes and TSB resources:

kubectl get pods -n bookinfo
NAME                                        READY   STATUS    RESTARTS   AGE
details-v1-79f774bdb9-8fr6d                 1/1     Running   0          4m17s
productpage-v1-6b746f74dc-mvl9n             1/1     Running   0          4m17s
ratings-v1-b6994bb9-zxq8n                   1/1     Running   0          4m17s
reviews-v1-545db77b95-c99dk                 1/1     Running   0          4m17s
reviews-v2-7bf8c9648f-rsndb                 1/1     Running   0          4m17s
reviews-v3-84779c7bbc-kzhwl                 1/1     Running   0          4m17s
tsb-gateway-bookinfo-73668b6aab-jygvk       1/1     Running   0          4m18s

kubectl get workspaces -A
NAMESPACE   NAME       PRIVILEGED   TENANT    AGE
bookinfo    bookinfo                tetrate   4m20s

# Use tctl to check for correct WS status. You can use the TSB UI instead.
$ tctl x status ws bookinfo
NAME        STATUS      LAST EVENT    MESSAGE
bookinfo    ACCEPTED

This means everything is up and running. The the bookinfo service can be accessed on the configured hostname through the ingress gateway.

If DNS is not configured in the cluster, or do you want to test it from your local environment, you can run a curl against the productpage service via its ingress gateway public IP like this.

$ export IP=$(kubectl -n bookinfo get service tsb-gateway-bookinfo -o jsonpath='{.status.loadBalancer.ingress[0].ip}')
$ curl -H "Host: bookinfo.example.com" http://$IP/productpage

Troubleshooting

Remember to bump the chart version when publishing new changes to the Chart.

If there are no changes and you want to force flux to re-run, do:

$ flux reconcile helmrelease bookinfo

You can also check for issues in the Flux Kubernetes resources:

$ flux get helmreleases -A
NAMESPACE   NAME        REVISION    SUSPENDED   READY   MESSAGE
flux-system bookinfo                False       False   install retries exhausted

kubectl get helmreleases -A -o yaml
...

If you see a upgrade retries exhausted message, there is a bug regression. The workaround is to suspend and resume the HelmRelease:

$ flux suspend helmrelease bookinfo
$ flux resume helmrelease bookinfo