Skip to content

Migration strategy

Before you begin the migration project, we suggest that you run through the exercises in Getting started to familiarize yourself with how Keptn works. When you are ready to begin the migration, follow the instructions in Installation and upgrade to set up your Kubernetes cluster and install Keptn on it.

On this page, we discuss some of the major paradigm shifts in Keptn relative to Keptn v1 and then discuss how elements of your Keptn v1 can be implemented for Keptn.

The Keptn cloud native paradigm

Keptn uses a different paradigm than Keptn v1 and so migration requires technical adjustments. Much of the logic and functionality of your Keptn v1 projects can be migrated to Keptn but must be rewritten to utilize Keptn components.

Keptn targets deployments that run on Kubernetes but some Keptn functionality can be used with deployments that do not run on Kubernetes. See Keptn for non-Kubernetes deployments for details.

Some key points:

  • Keptn uses native Kubernetes CRDs to configure SLIs/SLOs, tasks, and other elements that are part of the environment where the application deployment lives rather than using its own Git repo and its shipyard.yaml file as Keptn v1 does.

  • Keptn is not a delivery tool but it does provide a possibility to promote software to another stage in a staged environment which works with standard deployment tools such as ArgoCD, Flux, even kubectl apply. Keptn complements these tools by preventing the scheduling and deployment of workflows if the environment does not meet the user-defined requirements.

  • Keptn operates on a KeptnApp resource that is an amalgamation of multiple Kubernetes workloads, which together comprise the released application. Each workload deploys a separate microservice, which can be deployed at different times using different tools.

  • Keptn integrates with continuous delivery tools to insures that a KeptnApp is not deployed if it does not meet the user-defined requirements for all the associated workloads. It also exposes metrics to evaluate the success/status of a deployment.

  • Keptn provides an operator that can observe and orchestrate application-aware workload life cycles. This operator leverages Kubernetes webhooks and the Kubernetes scheduler to support pre- and post-deployment hooks. When the operator detects a new version of a service (implemented as a Kubernetes Workload), it can execute pre- and post-deployment evaluations and tasks using Kubernetes capabilities.

  • Keptn provides extensive observability data using OpenTelemetry and Prometheus rather than storing the data in a special Keptn database. This data can be displayed using Grafana and Jaeger or the dashboard of your choice.

For in-depth information about Keptn components and how they work, see the Architecture section.

Disposition of Keptn v1 components in Keptn

To help you wrap your mind around the migration process, this lists Keptn v1 components and identifies their possible disposition when you migrate to Keptn.


A Keptn v1 project represents an arbitrary, higher-level grouping of services. A project is defined by a shipyard.yaml file.

Keptn does not recognize projects. Many Keptn v1 projects may translate into KeptnApp resources but others will not. For example, if your Keptn v1 project has a large number of services, you may want to aggregate them into separate KeptnApp resources. A project has a 1:1 mapping to a Git repo, but, since few applications are stored in a mono-repo, it is difficult to implement a project-as-application paradigm.

In general, the Keptn v1 project is a useful as a reference list when migrating to Keptn. but may not directly translate to a Keptn resource.


A Keptn v1 service models a smaller chunk of a project. Most projects include many services. In a micro-services world, a service may represent a micro-service but it could instead be a wrapper for something else, like "the entire public website"

Keptn does not have the concept of a service. When migrating to Keptn, you need to analyze what each service is doing and translate that into an appropriate resource. The closest analogy is a Kubernetes workload but some services may be translated into KeptnTaskDefinition or other resources. See Working with Keptn tasks for more information.

For example:

  • A Keptn v1 service that runs chaos or load tests can probably be translated into a KeptnTask using the container-runner.
  • A Keptn v1 service that runs a database can probably be translated into a Kubernetes StateFulSet workload; see Workload Resources for more information.
  • A Keptn v1 service that runs a webserver can probably be translated into a Kubernetes Deployment workload.


A stage is a subsection of a project. Because Keptn is not a delivery tool, it has no concept of a stage but rather depends on a deployment engine. However, the logic of the stages can be useful when architecting the migration:

  • A deployment stage -- may define sequences of tasks that should be translated into KeptnTaskDefinition resources that are executed pre- and post-deployment
  • A testing stage may define sequences of tasks that should be translated into KeptnTaskDefinition resources that are executed pre- and post-deployment.

Stage functionality could be implemented in many different ways. Some functionality might be implemented in different namespaces or even in different Keptn-enabled clusters, allowing a tool such as ArgoCD to handle promotion.


A sequence is an ordered and user-defined sequence of tasks that are defined in a shipyard.yaml file. The shipyard controller micro-service reads the shipyard file and (when the time is right), emits a taskName.triggered cloud event onto the event bus. The shipyard controller waits to receive a taskName.started event and a correspondingly equal number of taskName.finished events before the shipyard controller reads the next task and emits a taskName.finished event for that task.

In this way, you can define arbitrary sequences of any tasks at any length and also link (or chain) sequences together to form (primitive) workflows. When migrating, these sequences of tasks can often be translated into KeptnTaskDefinition resources that are defined to run either pre- or post-deployment of the pod-scheduling phase.

The shipyard file is a general purpose workflow engine that is backed by cloud events. It is not opinionated to a tool, platform, technology, or a particular "slice" of the lifecycle. The TriggeredOn property allows a Keptn v1 sequence to be triggered at any time by a user or another system. This capability can be used, for example, to trigger a data encryption and backup operation, or a file movement over a network, or other arbitrary activities that may or may not have anything to do with an application's lifecycle.

When migrating to Keptn, sequences that are not part of the lifecycle workflow should not be handled by Keptn but should instead be handled by the pipeline engine tools being used such as Jenkins, Argo Workflows, Flux, and Tekton.


Keptn v1 defines some specific types of tasks, each of which is translated to a Keptn resource that is appropriate for the activity:

  • A Keptn v1 deployment task becomes a Deployment, StatefulSets, or DaemonSets, workload. You can code KeptnTaskDefinition and KeptnEvaluationDefinition resources that are configured to run either pre- or post-deployment tasks
  • An evaluation task becomes a KeptnEvaluationDefinition resource.
  • All other standard tasks (action, approval, get-action, rollback, release, test) as well as custom task types that might be defined should be translated into KeptnTaskDefinition resources.
  • The key:value properties for each Keptn v1 sequence should be coded into the KeptnTaskDefinition resource as appropriate.

Note that all KeptnTask resources at the same level (either pre-deployment or post-deployment) execute in parallel whereas Keptn v1 sequences and tasks can not execute in parallel.

  • If you have actions that need to execute sequentially, create a single KeptnTaskDefinition that calls each action in order.
  • If you have tasks that can execute in parallel, migrating to Keptn may improve the performance of the deployment.


Keptn v1 SLIs (Service Level Indicators) represent queries from the data provider such as Prometheus, Thanos, Cortex, Dynatrace, or Datadog, which is configured as a Keptn integration.

When migrating to Keptn, you need to define a KeptnMetricsProvider resource for the data provider(s) you are using. Note that Keptn allows you to support multiple data providers and multiple instances of each data provider for your SLIs whereas Keptn v1 only allows you to use one SLI per project.

The queries defined for the Keptn v1 SLIs should be translated into an appropriate Keptn resource:


Keptn v1 SLOs. can be implemented on Keptn as evaluations or analyses:


Keptn does not currently support the same level of remediations as Keptn v1 does, but it does provide limited "Day 2" facilities:

  • Any query that is possible for your data provider post-deployment can be defined as a KeptnMetricDefinition; this is then reported as a Keptn Metric. Evaluation can be defined as a KeptnEvaluationDefinition.
  • KeptnMetricsDefinition resources can be retrieved and used to implement the Kubernetes HorizontalPodAutoscaler (HPA), which can detect the need for additional resources (more pods, memory, disk space, etc.) and automatically add those resources to your configuration based on the ReplicaSet resources you have defined. See Using the HorizontalPodAutoscaler for more information.

Integrations and services in JES

Most functionality coded using the Keptn v1 JES (Job Executor Service) facility can simply be moved into a KeptnTaskDefinition resource that uses the container-runtime runner. If the JES container code is written in JavaScript or TypeScript, you may be able to use the deno-runtime runner. If the JES container code is written in Python 3, you may be able to use the python-runtime runner.

Note that there is no need for integrations for data providers in Keptn; these are configured as KeptnMetricsProvider resources.