Skip to main content

Kratix Workflow Controls: Easily Manual Approvals to Your Platform

· 6 min read
Cat Morris
Cat Morris
Product Manager @ Syntasso

Modern platforms do not operate in isolation. They depend on manual approvals, external systems, the availability of other resources, and compound building blocks. Until now, modelling these asynchronous processes in Kratix has been inelegant, and the underlying Kubernetes pods orchestrating the tasks could remain alive for a long time.

With the new Workflow Control feature, you can now safely and efficiently suspend a workflow until a condition is met, without leaving long-running pods hanging around. Your platform stays responsive, your workflows stay clean, and your enterprise processes fit naturally into Kratix.

Unlike CI pipelines or infrastructure tools that rely on polling or long-running jobs, Kratix natively understands when to pause, retry, and resume workflows as part of the platform.

Real platforms are asynchronous: most infra tools pretend they’re not​

Most Platform Engineers aspire to build platforms within a declarative world that requires no manual intervention, where dependencies don't prevent them from waiting for their desired states to converge... However, at Syntasso, we know that it isn't always possible, so we build for the enterprises that consistently face scenarios such as:

  • A database request must be manually approved in Jira before provisioning begins
  • One resource must wait until another resource is available
  • Creation of environments requires things to happen sequentially

Until now, the only way to model these behaviours in Kratix has been to wrap checks in loops and sleeps within pipeline containers.

This created two serious problems:

  1. Long-running pods consume resources unnecessarily
    • If approvals take days, those pods sit idle but still consume memory and, in some environments, hog scarce IP addresses.
  2. The pattern feels hacky and fragile
    • Promise writers are forced to build retry logic, polling, and lifecycle management themselves, rather than relying on Kratix to orchestrate workflow execution cleanly.

This left platform teams with an unnecessary cognitive overhead, not just in writing Promises but in how the whole platform behaves.

Workflow Control: native, efficient control of your pipelines’ behaviour​

Kratix managing jobs with Workflow Controls
Workflow Control in Action

The new Kratix Workflow Control functionality introduces a clean execution control mechanism that allows pipelines to:

  • Signal that a condition has not yet been met
  • Retry a specified number of times and after a defined interval
  • Suspend execution without being treated as a failure
  • Resumed when the condition is satisfied

Instead of keeping a pod alive on a sleep cycle, the pipeline exits and instructs Kratix when to retry. The pipeline still contains all the business-specific logic needed to determine whether the condition is met, while Kratix handles retrying the pipeline.

Name:         prod
Namespace:    default
API Version:  marketplace.kratix.io/v1alpha1
Kind:         app
Metadata:
  # ...
Spec:
  Image:                      syntasso/website
  Lifecycle:                  prod
Status:
  Approval Issue:  https://github.com/syntasso/demo-approvals/issues/131
  Conditions:
    Last Transition Time:  2026-04-09T13:15:56Z
    Message:               Pipelines are still in progress
    Reason:                PipelinesInProgress
    Status:                False
    Type:                  ConfigureWorkflowCompleted
    Last Transition Time:  2026-04-09T13:16:02Z
    Message:               Suspended
    Reason:                WorkflowSuspended
    Status:                Unknown
    Type:                  Reconciled
  Kratix:
    Workflows:
      Pipelines:
        Attempts:              1
        Last Transition Time:  2026-04-09T13:15:55Z
        Message:               Waiting for approval on issue https://github.com/syntasso/demo-approvals/issues/131
        Name:                  app-configure
        Next Retry At:         2026-04-09T13:16:32Z
        Phase:                 Suspended
      Suspended Generation:    1
  Message:                     Waiting for approval on issue https://github.com/syntasso/demo-approvals/issues/131
  Workflows:                   1
  Workflows Failed:            0
  Workflows Succeeded:         0
Events:
  Type     Reason             Age              From                       Message
  ----     ------             ----             ----                       -------
  Normal   PipelineStarted    11s              ResourceRequestController  Configure Pipeline started: app-configure
  Warning  WorkflowSuspended  2s (x6 over 5s)  ResourceRequestController  'kratix.io/workflow-suspended' label set to 'true' for resource request; skipping reconciliation

Example of a resource in the suspended state.

Using Workflow Controls to manage manual approvals​

In the video above, our Principal Engineer Derik shows how you can use this feature to manage manual approvals via Github issues.

Key Workflow Control Features​

  • ⏳ Efficient retry and suspend mechanism
    • Pipelines can create a workflow control file that instructs Kratix to retry after a specified interval, with optional max attempts, and the ability to suspend the checks if needed.
    • Save money by reducing the number of idle pods waiting for approvals
  • 🔁 Condition-aware progression
    • If a wait condition is not met, Kratix re-executes the pipeline after the configured delay. When the pipeline is marked “completed”, the next pipeline stage starts automatically.
    • This means no more manual checks on status, or if the pod is still running as you expect. Just check the status of your resource.
  • 🧠 Clear separation between failure and “retry later”
    • Non-zero exits are still treated as failures. But unmet conditions are explicitly handled as retry states, rather than errors
    • This makes workflow statuses clearer and easier to reason about.
  • 📊 Enhanced status visibility
    • Retry attempts, suspension state, and next retry timestamps are persisted in the resource status, making the wait behaviour transparent for users.
    • This reduces the need to poke another team to see what is going on under the platform-hood

What this means for platform teams​

With Workflow Controls, you can:

  • Integrate real-world change management workflows inside Kratix
  • Reduce platform resource pressure from idle pods
  • Avoid fragile sleep-and-poll scripts

Your platform no longer has to choose between correctness and efficiency.

Ready to try it out?​

Promise Pipeline Waits and Events are available now in the latest Kratix & SKE releases.

Real platforms are asynchronous. Now your Promise pipelines are too, without the operational cost.