Update gitops overview; add flux chapter

This commit is contained in:
Jérôme Petazzoni
2024-06-09 17:47:58 +02:00
parent ea16766fd7
commit 0d761409d7
4 changed files with 647 additions and 141 deletions

504
slides/k8s/flux.md Normal file
View File

@@ -0,0 +1,504 @@
# FluxCD
- We're going to implement a basic GitOps workflow with Flux
- Pushing to `main` will automatically deploy to the clusters
- There will be two clusters (`dev` and `prod`)
- The two clusters will have similar (but slightly different) workloads
---
## Repository structure
This is (approximately) what we're going to do:
```
@@INCLUDE[slides/k8s/gitopstree.txt]
```
---
## Getting ready
- Let's make sure we have two clusters
- It's OK to use local clusters (kind, minikube...)
- We might run into resource limits, though
(pay attention to `Pending` pods!)
- We need to install the Flux CLI ([packages], [binaries])
- **Highly recommended:** set up CLI completion!
- Of course we'll need a Git service, too
(we're going to use GitHub here)
[packages]: https://fluxcd.io/flux/get-started/
[binaries]: https://github.com/fluxcd/flux2/releases
---
## GitHub setup
- Generate a GitHub token:
https://github.com/settings/tokens/new
- Give it "repo" access
- This token will be used by the `flux bootstrap github` command later
- It will create a repository and configure it (SSH key...)
- The token can be revoked afterwards
---
## Flux bootstrap
.lab[
- Let's set a few variables for convenience, and create our repository:
```bash
export GITHUB_TOKEN=...
export GITHUB_USER=changeme
export GITHUB_REPO=alsochangeme
export FLUX_CLUSTER=dev
flux bootstrap github \
--owner=$GITHUB_USER \
--repository=$GITHUB_REPO \
--branch=main \
--path=./clusters/$FLUX_CLUSTER \
--personal --public
```
]
Problems? check next slide!
---
## What could go wrong?
- `flux bootstrap` will create or update the repository on GitHub
- Then it will install Flux controllers to our cluster
- Then it waits for these controllers to be up and running and ready
- Check pod status in `flux-system`
- If pods are `Pending`, check that you have enough resources on your cluster
- For testing purposes, it should be fine to lower or remove Flux `requests`!
(but don't do that in production!)
- If anything goes wrong, don't worry, we can just re-run the bootstrap
---
class: extra-details
## Idempotence
- It's OK to run that same `flux bootstrap` command multiple times!
- If the repository already exists, it will re-use it
(it won't destroy or empty it)
- If the path `./clusters/$FLUX_CLUSTER` already exists, it will update it
- It's totally fine to re-run `flux bootstrap` if something fails
- It's totally fine to run it multiple times on different clusters
- Or even to run it multiple times for the *same* cluster
(to reinstall Flux on that cluster after a cluster wipe / reinstall)
---
## What do we get?
- Let's look at what `flux bootstrap` installed on the cluster
.lab[
- Look inside the `flux-system` namespace:
```bash
kubectl get all --namespace flux-system
```
- Look at `kustomizations` custom resources:
```bash
kubectl get kustomizations --all-namespaces
```
- See what the `flux` CLI tells us:
```bash
flux get all
```
]
---
## Deploying with GitOps
- We'll need to add/edit files on the repository
- We can do it by using `git clone`, local edits, `git commit`, `git push`
- Or by editing online on the GitHub website
.lab[
- Create a manifest; for instance `clusters/dev/flux-system/blue.yaml`
- Add that manifest to `cluseters/dev/kustomization.yaml`
- Commit and push both changes to the repository
]
---
## Waiting for reconciliation
- Compare the git hash that we pushed and the one show with `kubectl get `
- Option 1: wait for Flux to pick up the changes in the repository
(the default interval for git repositories is 1 minute, so that's fast)
- Option 2: use `flux reconcile source git flux-system`
(this puts an annotation on the appropriate resource, triggering an immediate check)
---
## Checking progress
- `flux logs`
- `kubectl get gitrepositories --all-namespaces`
- `kubectl get kustomizations --all-namespaces`
---
## Did it work?
--
- No!
--
- Why?
--
- We need to indicate the namespace where the app should be deployed
- Either in the YAML manifests
- Or in the `kustomization` custom resource
(using field `spec.targetNamespace`)
- Add the namespace to the manifest and try again!
---
## Adding an app in a reusable way
- Let's see a technique to add a whole app
(with multiple resource manifets)
- We want that to be reusable
(i.e. easy to add on multiple clusters with minimal changes)
---
## The plan
- Add the app manifests in a directory
(e.g.: `apps/myappname/manifests`)
- Create a kustomization manifest for the app and its namespace
(e.g.: `apps/myappname/flux.yaml`)
- The kustomization manifest will refer to the app manifest
- Add the kustomization manifest to the top-level `flux-system` kustomization
---
## Creating the manifests
- All commands below should be executed at the root of the repository
.lab[
- Put application manifests in their directory:
```bash
mkdir -p apps/dockercoins
cp ~/container.training/k8s/dockercoins.yaml apps/dockercoins/
```
- Create kustomization manifest:
```bash
flux create kustomization dockercoins \
--source=GitRepository/flux-system \
--path=./apps/dockercoins/manifests/ \
--target-namespace=dockercoins \
--prune=true --export > apps/dockercoins/flux.yaml
```
]
---
## Creating the target namespace
- When deploying *helm releases*, it is possible to automatically create the namespace
- When deploying *kustomizations*, we need to create it explicitly
- Let's put the namespace with the kustomization manifest
(so that the whole app can be mediated through a single manifest)
.lab[
- Add the target namespace to the kustomization manifest:
```bash
echo "---
kind: Namespace
apiVersion: v1
metadata:
name: dockercoins" >> apps/dockercoins/flux.yaml
```
]
---
## Linking the kustomization manifest
- Edit `clusters/dev/flux-system/kustomization.yaml`
- Add a line to reference the kustomization manifest that we created:
```yaml
- ../../../apps/dockercoins/flux.yaml
```
- `git add` our manifests, `git commit`, `git push`
(check with `git status` that we haven't forgotten anything!)
- `flux reconcile` or wait for the changes to be picked up
---
## Installing with Helm
- We're going to see two different workflows:
- installing a third-party chart
<br/>
(e.g. something we found on the Artifact Hub)
- installing one of our own charts
<br/>
(e.g. a chart with authored ourselves)
- The procedures are very similar
---
## Installing from a public Helm repository
- Let's install [kube-prometheus-stack][kps]
.lab[
- Create the Flux manifests:
```bash
mkdir -p apps/kube-prometheus-stack
flux create source helm kube-prometheus-stack \
--url=https://prometheus-community.github.io/helm-charts \
--export >> apps/kube-prometheus-stack/flux.yaml
flux create helmrelease kube-prometheus-stack \
--source=HelmRepository/kube-prometheus-stack \
--chart=kube-prometheus-stack --release-name=kube-prometheus-stack \
--target-namespace=kube-prometheus-stack --create-target-namespace \
--export >> apps/kube-prometheus-stack/flux.yaml
```
]
[kps]: https://artifacthub.io/packages/helm/prometheus-community/kube-prometheus-stack
---
## Enable the app
- Just like before, link the manifest from the top-level kustomization
(`flux-system` in namespace `flux-system`)
- `git add` / `git commit` / `git push`
- We should now have a Prometheus+Grafana observability stack!
---
## Installing from a Helm chart in a git repo
- In this example, the chart will be in the same repo
- In the real world, it will typically be in a different repo!
.lab[
- Generate a basic Helm chart:
```bash
mkdir -p charts
helm create charts/myapp
```
]
(This generates a chart which installs NGINX. A lot of things can be customized, though.)
---
## Creating the Flux manifests
- The invocation is very similar to our first example
.lab[
- Generate the Flux manifest for the Helm release:
```bash
mkdir apps/myapp
flux create helmrelease myapp \
--source=GitRepository/flux-system \
--chart=charts/myapp \
--target-namespace=myapp --create-target-namespace \
--export > apps/myapp/flux.yaml
```
- Add that manifest to the top-level kustomization
- `git add` / `git commit` / `git push` the chart, manifest, and kustomization
]
---
## Passing values
- We can also configure our Helm releases with values
- Using an existing `myvalues.yaml` file:
`flux create helmrelease ... --values=myvalues.yaml`
- Referencing an existing ConfigMap or Secret with a `values.yaml` key:
`flux create helmrelease ... --values-from=ConfigMap/myapp`
---
## Gotchas
- When creating a HelmRelease using a chart stored in a git repository, you must:
- either bump the chart version (in `Chart.yaml`) after each change,
- or set `spec.chart.spec.reconcileStrategy` to `Revision`
- Why?
- Flux installs helm releases using packaged artifacts
- Artifacts are updated only when the Helm chart version changes
- Unless `reoncileStrategy` is set to `Revision` (instead of the default `ChartVersion`)
---
## More gotchas
- There is a bug in Flux that prevents using identical subcharts with aliases
- See [fluxcd/flux2#2505][flux2505] for details
[flux2505]: https://github.com/fluxcd/flux2/discussions/2505
---
## Things that we didn't talk about...
- Bucket sources
- Image automation controller
- Image reflector controller
- And more!
???
:EN:- Implementing gitops with Flux
:FR:- Workflow gitops avec Flux
<!--
helm upgrade --install --repo https://dl.gitea.io/charts --namespace gitea --create-namespace gitea gitea \
--set persistence.enabled=false \
--set redis-cluster.enabled=false \
--set postgresql-ha.enabled=false \
--set postgresql.enabled=true \
--set gitea.config.session.PROVIDER=db \
--set gitea.config.cache.ADAPTER=memory \
#
# Boostrap Flux controllers
```bash
mkdir -p flux/flux-system/gotk-components.yaml
flux install --export > flux/flux-system/gotk-components.yaml
kubectl apply -f flux/flux-system/gotk-components.yaml
```
# Bootstrap GitRepository/Kustomization
```bash
export REPO_URL="<gitlab_url>" DEPLOY_USERNAME="<username>"
read -s DEPLOY_TOKEN
flux create secret git flux-system --url="${REPO_URL}" --username="${DEPLOY_USERNAME}" --password="${DEPLOY_TOKEN}"
flux create source git flux-system --url=$REPO_URL --branch=main --secret-ref flux-system --ignore-paths='/*,!/flux' --export > flux/flux-system/gotk-sync.yaml
flux create kustomization flux-system --source=GitRepository/flux-system --path="./flux" --prune=true --export >> flux/flux-system/gotk-sync.yaml
git add flux/ && git commit -m 'feat: Setup Flux' flux/ && git push
kubectl apply -f flux/flux-system/gotk-sync.yaml
```
-->

13
slides/k8s/gitopstree.txt Normal file
View File

@@ -0,0 +1,13 @@
├── charts/ <--- could also be in separate app repos
│ ├── dockercoins/
│ └── color/
├── apps/ <--- YAML manifests for GitOps resources
│ ├── dockercoins/ (might reference the "charts" above,
│ ├── blue/ and/or include environment-specific
│ ├── green/ manifests to create e.g. namespaces,
│ ├── kube-prometheus-stack/ configmaps, secrets...)
│ ├── cert-manager/
│ └── traefik/
└── clusters/ <--- per-cluster; will typically reference
├── prod/ the "apps" above, possibly extending
└── dev/ or adding configuration resources too

View File

@@ -1,4 +1,4 @@
# Git-based workflows
# Git-based workflows (GitOps)
- Deploying with `kubectl` has downsides:
@@ -40,223 +40,210 @@
## Enabling git-based workflows
- There are a few tools out there to help us do that
- There are a many tools out there to help us do that
- We'll see demos of two of them: [Flux] and [Gitkube]
- There are *many* other tools, some of them with even more features
(examples: [ArgoCD], [FluxCD]...)
- There are also *many* integrations with popular CI/CD systems
(e.g.: GitLab, Jenkins, ...)
(e.g.: GitHub Actions, GitLab, Jenkins...)
[Flux]: https://www.weave.works/oss/flux/
[Gitkube]: https://gitkube.sh/
[ArgoCD]: https://argoproj.github.io/cd/
[Flux]: https://fluxcd.io/
---
## Flux overview
## The road to production
- We put our Kubernetes resources as YAML files in a git repository
In no specific order, we need to at least:
- Flux polls that repository regularly (every 5 minutes by default)
- Choose a tool
- The resources described by the YAML files are created/updated automatically
- Choose a cluster / app / namespace layout
<br/>
(one cluster per app, different clusters for prod/staging...)
- Changes are made by updating the code in the repository
- Choose a repository layout
<br/>
(different repositories, directories, branches per app, env, cluster...)
- Choose an installation / bootstrap method
- Choose how new apps / environments / versions will be deployed
- Choose how new images will be built
---
## Preparing a repository for Flux
## FluxCD
- We need a repository with Kubernetes YAML files
- Nice bootstrap
- I have one: https://github.com/jpetazzo/kubercoins
(CLI tool can automatically install, create git repos...)
- Fork it to your GitHub account
- Self-hosted
- Create a new branch in your fork; e.g. `prod`
(flux controllers are managed by flux itself)
(e.g. with "branch" dropdown through the GitHub web UI)
- Many CRDs
- This is the branch that we are going to use for deployment
(Kustomization, HelmRelease, GitRepository...)
- No web UI out of the box
- CLI relies on Kubernetes API access
---
## Setting up Flux with kustomize
## ArgoCD
- Clone the Flux repository:
```bash
git clone https://github.com/fluxcd/flux
cd flux
```
- Simple bootstrap
- Edit `deploy/flux-deployment.yaml`
(just apply YAMLs / install Helm chart)
- Change the `--git-url` and `--git-branch` parameters:
```yaml
- --git-url=git@github.com:your-git-username/kubercoins
- --git-branch=prod
```
- Few CRDs
- Apply all the YAML:
```bash
kubectl apply -k deploy/
```
(basic workflow can be done with a single "Application" resource)
- Comes with a web UI
- CLI relies on separate API and authentication system
---
## Setting up Flux with Helm
## Cluster, app, namespace layout
- Add Flux helm repo:
```bash
helm repo add fluxcd https://charts.fluxcd.io
```
- One cluster per app, different namespaces for environments?
- Install Flux:
```bash
kubectl create namespace flux
helm upgrade --install flux \
--set git.url=git@github.com:your-git-username/kubercoins \
--set git.branch=prod \
--namespace flux \
fluxcd/flux
```
- One cluster per environment, different namespaces for apps?
- Everything on a single cluster? One cluster per combination?
- Something in between:
- prod cluster, database cluster, dev/staging/etc cluster
- prod+db cluster per app, shared dev/staging/etc cluster
- And more!
Note: this decision isn't really tied to GitOps!
---
## Allowing Flux to access the repository
## Repository layout
- When it starts, Flux generates an SSH key
So many different possibilities!
- Display that key:
```bash
kubectl -n flux logs deployment/flux | grep identity.pub | cut -d '"' -f2
```
- Source repos
- Then add that key to the repository, giving it **write** access
- Cluster/infra repos/branches/directories
(some Flux features require write access)
- "Deployment" repos (with manifests, charts)
- After a minute or so, DockerCoins will be deployed to the current namespace
- Different repos/branches/directories for environments
🤔 How to decide?
---
## Making changes
## Permissions
- Make changes (on the `prod` branch), e.g. change `replicas` in `worker`
- Different teams/companies = different repos
- After a few minutes, the changes will be picked up by Flux and applied
- separate platform team → separate "infra" vs "apps" repos
- teams working on different apps → different repos per app
- Branches can be "protected" (`production`, `main`...)
(don't need separate repos for separate environments)
- Directories will typically have the same permissions
- Managing directories is easier than branches
- But branches are more "powerful" (cherrypicking, rebasing...)
---
## Other features
## Resource hierarchy
- Flux can keep a list of all the tags of all the images we're running
- Git-based deployments are managed by Kubernetes resources
- The `fluxctl` tool can show us if we're running the latest images
(e.g. Kustomization, HelmRelease with Flux; Application with ArgoCD)
- We can also "automate" a resource (i.e. automatically deploy new images)
- These resources need to be managed like any other Kubernetes resource
- And much more!
(YAML manifests, Kustomizations, Helm charts)
- They can be managed with Git workflows too!
---
## Gitkube overview
## Cluster / infra management
- We put our Kubernetes resources as YAML files in a git repository
- How do we provision clusters?
- Gitkube is a git server (or "git remote")
- Manual "one-shot" provisioning (CLI, web UI...)
- After making changes to the repository, we push to Gitkube
- Automation with Terraform, Ansible...
- Gitkube applies the resources to the cluster
- Kubernetes-driven systems (Crossplane, CAPI)
- Infrastructure can also be managed with GitOps
---
## Setting up Gitkube
## Example 1
- Install the CLI:
```bash
sudo curl -L -o /usr/local/bin/gitkube \
https://github.com/hasura/gitkube/releases/download/v0.2.1/gitkube_linux_amd64
sudo chmod +x /usr/local/bin/gitkube
```
- Managed with YAML/Charts:
- Install Gitkube on the cluster:
```bash
gitkube install --expose ClusterIP
```
- core components (CNI, CSI, Ingress, logging, monitoring...)
- GitOps controllers
- critical application foundations (database operator, databases)
- GitOps manifests
- Managed with GitOps:
- applications
- staging databases
---
## Creating a Remote
## Example 2
- Gitkube provides a new type of API resource: *Remote*
- Managed with YAML/Charts:
(this is using a mechanism called Custom Resource Definitions or CRD)
- essential components (CNI, CoreDNS)
- Create and apply a YAML file containing the following manifest:
```yaml
apiVersion: gitkube.sh/v1alpha1
kind: Remote
metadata:
name: example
spec:
authorizedKeys:
- `ssh-rsa AAA...`
manifests:
path: "."
```
- initial installation of GitOps controllers
(replace the `ssh-rsa AAA...` section with the content of `~/.ssh/id_rsa.pub`)
- Managed with GitOps:
- upgrades of GitOps controllers
- core components (CSI, Ingress, logging, monitoring...)
- operators, databases
- more GitOps manifests for applications!
---
## Pushing to our remote
## Concrete example
- Get the `gitkubed` IP address:
```bash
kubectl -n kube-system get svc gitkubed
IP=$(kubectl -n kube-system get svc gitkubed -o json |
jq -r .spec.clusterIP)
```
- Source code repository (not shown here)
- Get ourselves a sample repository with resource YAML files:
```bash
git clone git://github.com/jpetazzo/kubercoins
cd kubercoins
```
- Infrastructure repository (shown below), single branch
- Add the remote and push to it:
```bash
git remote add k8s ssh://default-example@$IP/~/git/default-example
git push k8s master
```
---
## Making changes
- Edit a local file
- Commit
- Push!
- Make sure that you push to the `k8s` remote
---
## Other features
- Gitkube can also build container images for us
(see the [documentation](https://github.com/hasura/gitkube/blob/master/docs/remote.md) for more details)
- Gitkube can also deploy Helm charts
(instead of raw YAML files)
```
@@INCLUDE[slides/k8s/gitopstree.txt]
```
???

View File

@@ -124,6 +124,9 @@ content:
- k8s/portworx.md
- k8s/openebs.md
- k8s/stateful-failover.md
-
- k8s/gitworkflows.md
- k8s/flux.md
-
- k8s/logs-centralized.md
- k8s/prometheus.md
@@ -164,7 +167,6 @@ content:
- k8s/cluster-upgrade.md
- k8s/cluster-backup.md
- k8s/cloud-controller-manager.md
- k8s/gitworkflows.md
-
- k8s/lastwords.md
- k8s/links.md