GitOps Deployment to AWS EKS with Kestra, Kaniko and ECR

id: aws-eks-gitops-deploy
namespace: company.team
description: |
  Deploy a containerized application to AWS EKS on every Git push: clone the
  repository into a shared working directory, build the image with Kaniko (no
  Docker daemon) and push it to Amazon ECR, apply the Kubernetes manifests,
  wait for the rolling update to finish, and report the result to Discord.
  If any step fails, capture pod status, logs and cluster events and alert
  Discord with that evidence.

concurrency:
  behavior: QUEUE
  limit: 1

inputs:
  - id: git_repo
    type: STRING
    defaults: "https://github.com/your-username/eks-gitops-deploy.git"
    description: HTTPS URL of the Git repository to build and deploy. It must
      contain a Dockerfile and Kubernetes manifests in k8s/.

  - id: git_branch
    type: STRING
    defaults: "main"
    description: Branch to check out, build and deploy.

  - id: aws_region
    type: STRING
    defaults: "ap-south-2"
    description: AWS region where the ECR repository and the EKS cluster live.

  - id: ecr_registry
    type: STRING
    defaults: "your-account-id.dkr.ecr.ap-south-2.amazonaws.com"
    description: ECR registry host, in the form
      <account-id>.dkr.ecr.<region>.amazonaws.com. The image is pushed to
      <registry>/eks-gitops-deploy:<execution id>.

  - id: eks_cluster_name
    type: STRING
    defaults: "eks-gitops-deploy-cluster"
    description: Name of the EKS cluster to deploy to. It is used to generate the
      kubeconfig with aws eks update-kubeconfig.

  - id: k8s_namespace
    type: STRING
    defaults: "production"
    description: Kubernetes namespace to deploy into. It is created if it does not
      exist, and the namespace in the k8s/ manifests is rewritten to match.

  - id: notify_discord
    type: BOOL
    defaults: true
    description: Post the success embed to Discord after a rollout. Requires the
      DISCORD_WEBHOOK_URL secret or KV pair; set to false to skip it. Failures
      are always reported by the error handler.

tasks:
  - id: working_dir
    type: io.kestra.plugin.core.flow.WorkingDirectory
    description: One shared directory so the clone, the Kaniko build and the kubectl
      deployment all see the same files, including the manifest edited with the
      new image tag.
    tasks:
      - id: clone_repository
        type: io.kestra.plugin.git.Clone
        description: Check out the repository at the configured branch.
        url: "{{ inputs.git_repo | trim }}"
        branch: "{{ inputs.git_branch | trim }}"

      - id: build_and_push_image
        type: io.kestra.plugin.scripts.shell.Commands
        description: Build the Dockerfile with the Kaniko executor in userspace (no
          privileged Docker daemon or mounted docker.sock) and push it to ECR,
          tagged with the execution ID so every deployment maps to exactly one
          run.
        taskRunner:
          type: io.kestra.plugin.scripts.runner.docker.Docker
        containerImage: gcr.io/kaniko-project/executor:debug
        env:
          AWS_ACCESS_KEY_ID: "{{ kv(key='AWS_ACCESS_KEY_ID', errorOnMissing=false) ??
            secret('AWS_ACCESS_KEY_ID') }}"
          AWS_SECRET_ACCESS_KEY: "{{ kv(key='AWS_SECRET_ACCESS_KEY', errorOnMissing=false)
            ?? secret('AWS_SECRET_ACCESS_KEY') }}"
          AWS_REGION: "{{ inputs.aws_region | trim }}"
        commands:
          - /kaniko/executor --context dir://. --dockerfile Dockerfile
            --destination {{ inputs.ecr_registry | trim }}/eks-gitops-deploy:{{
            execution.id }} --custom-platform=linux/amd64 --ignore-path=/tmp

      - id: deploy_and_verify_rollout
        type: io.kestra.plugin.scripts.shell.Commands
        description: Connect to EKS, inject the new image tag into k8s/deployment.yaml,
          apply the manifests, block on kubectl rollout status until the new
          pods are ready, then write the Discord success embed (LoadBalancer
          URL, pod placement and replica count) to a file for the notification
          step.
        taskRunner:
          type: io.kestra.plugin.scripts.runner.docker.Docker
        containerImage: alpine/k8s:1.31.0
        env:
          AWS_ACCESS_KEY_ID: "{{ kv(key='AWS_ACCESS_KEY_ID', errorOnMissing=false) ??
            secret('AWS_ACCESS_KEY_ID') }}"
          AWS_SECRET_ACCESS_KEY: "{{ kv(key='AWS_SECRET_ACCESS_KEY', errorOnMissing=false)
            ?? secret('AWS_SECRET_ACCESS_KEY') }}"
          AWS_DEFAULT_REGION: "{{ inputs.aws_region | trim }}"
        outputFiles:
          - discord_success.json
        commands:
          - aws eks update-kubeconfig --region {{ inputs.aws_region | trim }}
            --name {{ inputs.eks_cluster_name | trim }}
          - kubectl create namespace {{ inputs.k8s_namespace | trim }}
            --dry-run=client -o yaml | kubectl apply -f -
          - sed -i "s|IMAGE_PLACEHOLDER|{{ inputs.ecr_registry | trim
            }}/eks-gitops-deploy:{{ execution.id }}|g" k8s/deployment.yaml
          - 'sed -i "s|namespace: production|namespace: {{ inputs.k8s_namespace
            | trim }}|g" k8s/*.yaml'
          - kubectl apply -f k8s/
          - kubectl rollout status deployment/eks-gitops-deploy -n {{
            inputs.k8s_namespace | trim }} --timeout=420s
          - |
            LB_HOST=$(kubectl get svc eks-gitops-deploy-service -n {{ inputs.k8s_namespace | trim }} -o jsonpath='{.status.loadBalancer.ingress[0].hostname}' 2>/dev/null || echo "")
            if [ -z "$LB_HOST" ]; then
              LB_HOST=$(kubectl get svc eks-gitops-deploy-service -n {{ inputs.k8s_namespace | trim }} -o jsonpath='{.status.loadBalancer.ingress[0].ip}' 2>/dev/null || echo "Pending...")
            fi
            POD_SUMMARY=$(kubectl get pods -n {{ inputs.k8s_namespace | trim }} -l app=eks-gitops-deploy -o wide --no-headers 2>/dev/null | awk '{print "• "$1" ("$3", Ready: "$2") on Node: "$7}' | head -n 5)
            REPLICAS=$(kubectl get deployment eks-gitops-deploy -n {{ inputs.k8s_namespace | trim }} -o jsonpath='{.status.readyReplicas}/{.spec.replicas}' 2>/dev/null || echo "2/2")

            jq -n \
              --arg cluster "{{ inputs.eks_cluster_name }}" \
              --arg region "{{ inputs.aws_region }}" \
              --arg ns "{{ inputs.k8s_namespace | trim }}" \
              --arg img "{{ inputs.ecr_registry }}/eks-gitops-deploy:{{ execution.id }}" \
              --arg execId "{{ flow.id }} - {{ execution.id }}" \
              --arg lb "http://$LB_HOST" \
              --arg pods "$POD_SUMMARY" \
              --arg reps "$REPLICAS Ready" \
              '{
                username: "eks-gitops-deploy Bot",
                avatar_url: "https://kestra.io/logo.png",
                embeds: [
                  {
                    title: "🚀 AWS EKS Deployment SUCCESS",
                    description: "**Application**: `eks-gitops-deploy` successfully deployed and rolled out to **{{ inputs.k8s_namespace | trim }}** on AWS EKS.",
                    color: 3066993,
                    fields: [
                      {
                        name: "🌐 Live Service LoadBalancer URL",
                        value: $lb,
                        inline: false
                      },
                      {
                        name: "📦 Container Image",
                        value: ("`" + $img + "`"),
                        inline: false
                      },
                      {
                        name: "🎯 Infrastructure Details",
                        value: ("• **Cluster**: `" + $cluster + "`\n• **Namespace**: `" + $ns + "`\n• **Region**: `" + $region + "`"),
                        inline: true
                      },
                      {
                        name: "📊 Pod Health & Sizing",
                        value: ("• **Replicas**: `" + $reps + "`\n• **CPU/RAM**: `100m-300m / 160Mi-320Mi`\n• **Probes**: Liveness & Readiness UP"),
                        inline: true
                      },
                      {
                        name: "📋 Active Pod Instances",
                        value: ("```text\n" + (if $pods != "" then $pods else "All pods healthy" end) + "\n```"),
                        inline: false
                      },
                      {
                        name: "🔗 Telemetry Endpoints Hub",
                        value: ("[Discovery Map](" + $lb + "/) • [Live System Stats](" + $lb + "/api/system) • [Actuator Health](" + $lb + "/actuator/health) • [App Info](" + $lb + "/api/info)"),
                        inline: false
                      }
                    ],
                    footer: {
                      text: ("Kestra GitOps Orchestrator • Execution: " + $execId)
                    },
                    timestamp: (now | todate)
                  }
                ]
              }' > discord_success.json

  - id: check_discord_enabled
    type: io.kestra.plugin.core.flow.If
    description: Only post the success embed when notifications are enabled, so the
      flow also works for teams that have no Discord webhook.
    condition: "{{ inputs.notify_discord }}"
    then:
      - id: send_success_embed
        type: io.kestra.plugin.scripts.shell.Commands
        description: POST the success embed built by the deploy task (LoadBalancer URL,
          image, replicas and pod placement) to the Discord webhook.
        taskRunner:
          type: io.kestra.plugin.scripts.runner.docker.Docker
        containerImage: alpine/k8s:1.31.0
        inputFiles:
          discord_success.json: "{{
            outputs.deploy_and_verify_rollout.outputFiles['discord_success.json\
            '] }}"
        env:
          DISCORD_WEBHOOK_URL: "{{ kv(key='DISCORD_WEBHOOK_URL', errorOnMissing=false) ??
            secret('DISCORD_WEBHOOK_URL') }}"
        commands:
          - 'curl -s -H "Content-Type: application/json" -X POST -d
            @discord_success.json "$DISCORD_WEBHOOK_URL"'
    else:
      - id: log_notifications_disabled
        type: io.kestra.plugin.core.log.Log
        description: Record that the Discord notification was skipped on purpose.
        message: "Deployment of {{ inputs.ecr_registry | trim }}/eks-gitops-deploy:{{
          execution.id }} finished. Discord notifications are disabled for this
          run."

# Error handler
errors:
  - id: capture_pod_diagnostics_and_alert
    type: io.kestra.plugin.scripts.shell.Commands
    description: When any task fails, collect pod status, recent container logs and
      cluster events and send them to Discord, because a failed rollout is only
      useful to the team if the evidence arrives with the alert.
    taskRunner:
      type: io.kestra.plugin.scripts.runner.docker.Docker
    containerImage: alpine/k8s:1.31.0
    env:
      AWS_ACCESS_KEY_ID: "{{ kv(key='AWS_ACCESS_KEY_ID', errorOnMissing=false) ??
        secret('AWS_ACCESS_KEY_ID') }}"
      AWS_SECRET_ACCESS_KEY: "{{ kv(key='AWS_SECRET_ACCESS_KEY', errorOnMissing=false)
        ?? secret('AWS_SECRET_ACCESS_KEY') }}"
      AWS_DEFAULT_REGION: "{{ inputs.aws_region | trim }}"
      DISCORD_WEBHOOK_URL: "{{ kv(key='DISCORD_WEBHOOK_URL', errorOnMissing=false) ??
        secret('DISCORD_WEBHOOK_URL') }}"
    commands:
      - |
        set +e
        aws eks update-kubeconfig --region {{ inputs.aws_region | trim }} --name {{ inputs.eks_cluster_name | trim }} || true
        echo "=== CAPTURING DEPLOYMENT DIAGNOSTICS ==="
        POD_STATUS=$(kubectl get pods -n {{ inputs.k8s_namespace | trim }} -l app=eks-gitops-deploy 2>&1 | head -n 6 | head -c 800)
        CRASH_LOGS=$(kubectl logs deployment/eks-gitops-deploy -n {{ inputs.k8s_namespace | trim }} --tail=15 --all-containers=true 2>&1 | head -c 800)
        EVENTS=$(kubectl get events -n {{ inputs.k8s_namespace | trim }} --sort-by='.metadata.creationTimestamp' 2>&1 | tail -n 6 | head -c 800)

        [ -z "$POD_STATUS" ] && POD_STATUS="No pod status available"
        [ -z "$CRASH_LOGS" ] && CRASH_LOGS="No container logs available"
        [ -z "$EVENTS" ] && EVENTS="No recent events"

        jq -n \
          --arg cluster "{{ inputs.eks_cluster_name }}" \
          --arg region "{{ inputs.aws_region }}" \
          --arg ns "{{ inputs.k8s_namespace | trim }}" \
          --arg img "{{ inputs.ecr_registry }}/eks-gitops-deploy:{{ execution.id }}" \
          --arg execId "{{ flow.id }} - {{ execution.id }}" \
          --arg podStatus "$POD_STATUS" \
          --arg crashLogs "$CRASH_LOGS" \
          --arg events "$EVENTS" \
          '{
            username: "eks-gitops-deploy Alert",
            avatar_url: "https://kestra.io/logo.png",
            embeds: [
              {
                title: "🚨 AWS EKS Deployment FAILED",
                description: ("Deployment failed during verification/rollout on cluster **" + $cluster + "**."),
                color: 15158332,
                fields: [
                  {
                    name: "🎯 Infrastructure & Context",
                    value: ("• **Cluster**: `" + $cluster + "`\n• **Namespace**: `" + $ns + "`\n• **Region**: `" + $region + "`\n• **Attempted Image**: `" + $img + "`"),
                    inline: false
                  },
                  {
                    name: "📋 Pod Health & Diagnostics",
                    value: ("```text\n" + $podStatus + "\n```"),
                    inline: false
                  },
                  {
                    name: "📜 Recent Crash / Error Logs",
                    value: ("```text\n" + $crashLogs + "\n```"),
                    inline: false
                  },
                  {
                    name: "⚠️ Cluster Events",
                    value: ("```text\n" + $events + "\n```"),
                    inline: false
                  }
                ],
                footer: {
                  text: ("Kestra GitOps Orchestrator • Execution: " + $execId)
                },
                timestamp: (now | todate)
              }
            ]
          }' > /tmp/discord_failure.json

        if [ -n "$DISCORD_WEBHOOK_URL" ]; then
          echo "Sending incident embed to Discord..."
          curl -s -H "Content-Type: application/json" -X POST -d @/tmp/discord_failure.json "$DISCORD_WEBHOOK_URL"
        else
          echo "Warning: DISCORD_WEBHOOK_URL is not set!"
        fi

triggers:
  - id: github_webhook
    type: io.kestra.plugin.core.trigger.Webhook
    description: GitHub (or your CI) POSTs here on every push to start a deployment.
      The URL is keyed, so treat the key as a secret.
    key: "{{ kv(key='DEPLOY_WEBHOOK_KEY', errorOnMissing=false) ??
      secret('DEPLOY_WEBHOOK_KEY') }}"

outputs:
  - id: deployed_image
    type: STRING
    description: Full image reference that was built, pushed to ECR and deployed,
      e.g.
      123456789012.dkr.ecr.ap-south-2.amazonaws.com/eks-gitops-deploy:3Fz8Qk1xYvLm.
    value: "{{ inputs.ecr_registry | trim }}/eks-gitops-deploy:{{ execution.id }}"
  - id: deployed_namespace
    type: STRING
    description: Kubernetes namespace the application was deployed into, e.g. production.
    value: "{{ inputs.k8s_namespace | trim }}"

Turn a Git push into a verified AWS EKS deployment without a CI server or a privileged Docker daemon. The flow builds the image with Kaniko in userspace, pushes it to Amazon ECR with an immutable per-execution tag, applies your manifests, waits for the rolling update and reports the outcome to Discord.

How it works

The github_webhook trigger starts the flow on every push, and concurrency with QUEUE runs rapid pushes one at a time.
clone_repository checks out git_repo at git_branch inside a WorkingDirectory.
build_and_push_image runs the Kaniko executor and pushes <ecr_registry>/eks-gitops-deploy:<execution.id> to ECR.
deploy_and_verify_rollout connects to EKS, injects the image tag and namespace into k8s/, applies the manifests and blocks on kubectl rollout status.
check_discord_enabled (If) posts the success embed when the notify_discord input is true, otherwise it logs that it was skipped.
If any task fails, the errors handler sends pod status, logs and cluster events to Discord.

What you get

Daemonless, unprivileged image builds.
One immutable image tag per execution, so each deployment maps to one Kestra run.
Discord alerts with the LoadBalancer URL on success and pod logs and events on failure.

Who it's for

Platform and DevOps engineers who want GitOps-style deploys on AWS EKS without a separate CI server.

Why orchestrate this with Kestra

A shell script can run the same commands, but it cannot queue overlapping pushes, keep a per-run history or route failures to a dedicated handler.

Prerequisites

An EKS cluster, an ECR repository named eks-gitops-deploy, and a Git repository with a Dockerfile and k8s/ manifests whose Deployment image is IMAGE_PLACEHOLDER.
A Kestra instance where the Docker task runner can run containers.

Secrets

Each can be a Kestra Secret or a KV pair.

AWS_ACCESS_KEY_ID and AWS_SECRET_ACCESS_KEY: credentials with ECR push and EKS access.
DEPLOY_WEBHOOK_KEY: any random string used as the last segment of the webhook URL.
DISCORD_WEBHOOK_URL: needed only when notify_discord is true.

Quick start

Add the secrets above.
Execute the flow with your git_repo, ecr_registry (host only), eks_cluster_name and aws_region.
Optionally point a GitHub push webhook at the flow's webhook URL.

How to extend

Add a Parallel task to smoke-test the new LoadBalancer URL, or swap Discord for Slack with the matching plugin.

GitOps CI/CD to AWS EKS with Kaniko, ECR and Discord Alerts

How it works

What you get

Who it's for

Why orchestrate this with Kestra

Prerequisites

Secrets

Quick start

How to extend

Links

Tasks

Orchestrate with Kestra

Share this Blueprint

Create a Blueprint

New to Kestra?

GitOps CI/CD to AWS EKS with Kaniko, ECR and Discord Alerts

How it works

What you get

Who it's for

Why orchestrate this with Kestra

Prerequisites

Secrets

Quick start

How to extend

Links

Tasks

Orchestrate with Kestra

Share this Blueprint

Create a Blueprint

More Cloud Blueprints

Descheduler Pod Eviction Threshold and Node Imbalance Sentinel

Weekly DigitalOcean Account Inventory Report

Delta Lake Uncompacted Small Files Compactor and Safe Vacuum Gate

Ephemeral DigitalOcean Droplet Runner with Guaranteed Teardown

Blue-Green DNS Cutover on DigitalOcean Domains

DigitalOcean Account Balance Depletion Sentinel

More Infrastructure Blueprints

Verify an npm dependency upgrade against the project's own tests before it is merged

Continuous GenAI Evaluation and Hallucination Regression Gate with DeepEval

Gate a Lockfile on Licence Policy and Known Advisories with deps.dev

DeploymentWatch - Kubernetes stuck rollout watcher

Hold Production Deploys While a Freeze Window Is Active

Triage Dependabot pull requests and alert Slack only on risky updates

New to Kestra?