Executions

An Execution is a running instance of a Function. When you invoke a Function, kontinue creates an Execution resource that tracks the workflow’s progress and stores its durable state so it can be resumed after failures.

Each Execution is a Kubernetes custom resource that persists the function’s state, child references, and intermediate results. If a worker crashes or is restarted, the Execution status allows the worker (or a new worker) to pick up from the last known state.

Creating an Execution

There are several ways to create an Execution:

Using the CLI

kontinue execute hello-world --interactive

Using a Kubernetes manifest

apiVersion: kontinue.cloud/v1alpha1
kind: Execution
metadata:
  generateName: hello-world-
spec:
  function: hello-world
  arguments:
    name: "Alice"

Apply with:

kubectl create -f execution.yaml

Using the kontinue SDK

exec, err := cli.SpawnExecute(ctx, "hello-world", &HelloWorldArgs{
    Name: "Alice",
}, &client.SpawnOptions{})

Execution Phases

An Execution progresses through several phases during its lifecycle:

Phase	Description
`Pending`	Execution created, waiting for a worker to pick it up
`Executing`	A worker is actively running the function
`Suspended`	Execution is waiting on a child resource (sub-execution, job, or suspension)
`Backoff`	Execution failed and is waiting before retry
`Completed`	Function returned successfully
`Failed`	Function returned an error (after exhausting retries)
`Canceled`	Execution was externally cancelled (e.g. via API)

Execution Spec

The spec defines what function to run and how:

spec:
  function: deploy-cluster      # Required: name of the registered function
  arguments:                    # Optional: input arguments passed to the function
    environment: production
    replicas: 3
  retry:                        # Optional: retry configuration
    retries: 3                  # Number of retry attempts (default: 0)
    backoff: 10s                # Wait time between retries (default: 5s)
    childPolicy: RetryFailed    # How to handle children on retry
  timeout:                      # Optional: timeout configuration
    overall: 1h                 # Max total time across all attempts
    attempt: 15m                # Max time per attempt
  ttl:                          # Optional: cleanup configuration
    deleteAfter: 24h            # Auto-delete after completion

Execution Status

The status field tracks the current state:

status:
  phase: Executing
  message: "Deploying to production cluster"
  workerName: worker-abc123
  startedAt: "2025-01-15T10:30:00Z"
  attempts:
    - phase: Failed
      message: "connection timeout"
      startedAt: "2025-01-15T10:25:00Z"
      finishedAt: "2025-01-15T10:26:00Z"
    - phase: Executing
      startedAt: "2025-01-15T10:30:00Z"
  progress:
    current: 2
    max: 5

Monitoring Executions

List all executions:

kontinue list
# or
kubectl get executions

Get execution details:

kontinue show <execution-name>
# or
kubectl describe execution <execution-name>