Live Migrate Production Workloads from OpenShift Templates to Helm Charts

Table of Contents

Introduction

In an earlier post, we looked at how to convert OpenShift templates to helm charts. In this post, we’ll see how to live migrate running production customer workloads from being managed by OpenShift templates to Helm charts.

The Challenge

Helm is not designed to be used as a target of migration from another deployment-management system like OCP templates. Helm assumes that when Helm charts are installed, they will create afresh all Kubernetes objects included in the chart.

In our case, many Kubernetes objects, that are now packaged as Helm charts, were already created in the OCP cluster using OCP templates. When you install a chart, it simply tries to create all objects & fails because objects by the same name already exist in the cluster.

This document explores a workaround for this issue.

The Workaround

Helm uses a label & 2 annotations to keep track of which objects in the cluster are managed by Helm & which Helm release created them. Specifically, every Helm-managed object has:

metadata:
  name: ...
  labels:
    app.kubernetes.io/managed-by: Helm
  annotations:
    meta.helm.sh/release-name: ...
    meta.helm.sh/release-namespace: ...

It’s possible to make Helm β€œadopt” existing cluster objects by adding these labels & annotations manually.

The steps are as follows:

  1. Create an β€œempty” Helm release of the Helm chart that contains objects you wish to adopt
  2. Label & annotate all objects
  3. Run helm upgrade

Helm has now adopted existing cluster objects! All subsequent helm upgrades will work as if these objects were created by Helm itself!

Let us now explore this workaround’s implementation using a simple example.

Existing ConfigMap

Say you have a ConfigMap deployed in a cluster:

apiVersion: v1
data:
  key: value
kind: ConfigMap
metadata:
  name: my-cm
  namespace: my-ns

Existing Helm Chart

Say you also have a Helm chart with the same ConfigMap:

tree my-chart
my-chart
β”œβ”€β”€ Chart.yaml
└── templates
    └── config-map.yaml
cat my-chart/templates/config-map.yaml
apiVersion: v1
data:
  key: value
kind: ConfigMap
metadata:
  name: my-cm

Try Installing the Helm Chart

Now, if you try to install this chart, you get an error:

helm install my-release my-chart
Error: INSTALLATION FAILED:
rendered manifests contain a resource that already exists.
Unable to continue with install:
ConfigMap "my-cm" in namespace "my-ns" exists and cannot be imported into the current release:
invalid ownership metadata;
label validation error: missing key "app.kubernetes.io/managed-by": must be set to "Helm";
annotation validation error: missing key "meta.helm.sh/release-name": must be set to "my-release";
annotation validation error: missing key "meta.helm.sh/release-namespace": must be set to "my-ns"

Add Label & Annotations

The last 3 lines of the error above, tell us exactly what we need to do to make our ConfigMap adoptable by Helm. So let’s do that:

$ kubectl label cm my-cm app.kubernetes.io/managed-by=Helm
configmap/my-cm labeled
$ kubectl annotate cm my-cm meta.helm.sh/release-name=my-release
configmap/my-cm annotated
$ kubectl annotate cm my-cm meta.helm.sh/release-namespace=helm-test
configmap/my-cm annotated

The ConfigMap now looks like this:

kubectl get cm my-cm -o yaml
apiVersion: v1
data:
  key: value
kind: ConfigMap
metadata:
  annotations:
    meta.helm.sh/release-name: my-release
    meta.helm.sh/release-namespace: my-ns
  labels:
    app.kubernetes.io/managed-by: Helm
  name: my-cm
  namespace: my-ns

At this point, even though the required labels & annotations exist on the object, it’s not being managed by Helm at all because there exists no Helm release in the cluster that can adopt this ConfigMap resource.

Create an Empty Helm Release

So the next step is to create an β€œempty” Helm release. It has to be empty in the sense that the chart cannot contain any objects. If it does, we will get the same error as before.

So first, create an empty chart:

$ tree my-chart
my-chart
└── Chart.yaml
$ cat my-chart/Chart.yaml
apiVersion: v2
name: my-chart
type: application
version: 0.1.0

Now, install this chart:

$ helm list | grep my-chart
<no output>
$ helm install my-release my-chart
NAME: my-release
NAMESPACE: my-ns
STATUS: deployed
REVISION: 1
TEST SUITE: None

At this point, even though the chart is installed, it hasn’t yet adopted our ConfigMap. You can verify this by running:

$ helm get all my-release

NAME: my-release
NAMESPACE: my-ns
STATUS: deployed
REVISION: 1
TEST SUITE: None

USER-SUPPLIED VALUES: null
COMPUTED VALUES: {}
HOOKS:
MANIFEST:

Adopt Existing Objects

The next step is to use the original Helm chart, with the ConfigMap object to finally adopt the ConfigMap that already exists in the cluster. We do this by running helm upgrade using the chart:

$ tree my-chart

my-chart
β”œβ”€β”€ Chart.yaml
└── templates
    └── config-map.yaml
$ helm upgrade my-release my-chart

Release "my-release" has been upgraded. Happy Helming!
NAME: my-release
NAMESPACE: my-ns
STATUS: deployed
REVISION: 2
TEST SUITE: None

This is when the cluster objects are actually adopted by Helm. We can verify this:

$ helm get all my-release

NAME: my-release
NAMESPACE: my-ns
STATUS: deployed
REVISION: 2
TEST SUITE: None

USER-SUPPLIED VALUES: null
COMPUTED VALUES: {}
HOOKS:

MANIFEST:
# Source: my-chart/templates/config-map.yaml
apiVersion: v1
data:
  key: value
kind: ConfigMap
metadata:
  name: my-cm

Helm Upgrade

From this point on, helm upgrade will work as expected. Let’s try this by adding some data to the ConfigMap:

apiVersion: v1
data:
  key: value
  new-key: new-value
kind: ConfigMap
metadata:
  name: my-cm
$ helm upgrade my-release my-chart

Release "my-release" has been upgraded. Happy Helming!
NAME: my-release
NAMESPACE: my-ns
STATUS: deployed
REVISION: 3
TEST SUITE: None
$ kubectl get cm my-cm -o yaml

apiVersion: v1
data:
  key: value
  new-key: new-value
kind: ConfigMap
metadata:
  annotations:
    meta.helm.sh/release-name: my-release
    meta.helm.sh/release-namespace: my-ns
  labels:
    app.kubernetes.io/managed-by: Helm
  name: my-cm

Helm Uninstall

As the objects are now managed by Helm, helm uninstall will delete them as well:

$ helm uninstall my-release
release "my-release" uninstalled

$ kubectl get cm my-cm
Error from server (NotFound): configmaps "my-cm" not found

Conclusion

This article was quite an elaborate walkthrough of how you can accomplish safe & zero-downtime migration of workloads managed by OpenShift templates to Helm charts. Head on over to part 3 of this blog series to learn how to migrate DeploymentConfigs to Deployments & Routes to Ingresses.

References

How to apply helm chart to existing resources?

How to Include New Kubernetes Resource Into Existing Helm Release

About the Author ✍🏻

Harish KM is a Principal DevOps Engineer at QloudX & a top-ranked AWS Ambassador since 2020. πŸ‘¨πŸ»β€πŸ’»

With over a decade of industry experience as everything from a full-stack engineer to a cloud architect, Harish has built many world-class solutions for clients around the world! πŸ‘·πŸ»β€β™‚οΈ

With over 20 certifications in cloud (AWS, Azure, GCP), containers (Kubernetes, Docker) & DevOps (Terraform, Ansible, Jenkins), Harish is an expert in a multitude of technologies. πŸ“š

These days, his focus is on the fascinating world of DevOps & how it can transform the way we do things! πŸš€

One Reply to “Live Migrate Production Workloads from OpenShift Templates to Helm Charts”

  1. […] Live Migrate Production Workloads from OpenShift Templates to Helm Charts […]

Leave a Reply

Your email address will not be published. Required fields are marked *