Skip to main content

Verify Installation

After deploying Yardrail, walk through these checks to confirm the system is working end-to-end.

1. Check running pods​

kubectl -n yardrail get pods

Both the frontend and worker pods should be Running:

NAME READY STATUS RESTARTS AGE
yardrail-frontend-7b8f9d4c5-x2k4m 1/1 Running 0 2m
yardrail-worker-5d6f8a3b2-p9n7j 1/1 Running 0 2m

2. Check the frontend​

The frontend should be reachable at your configured ingress or service:

kubectl -n yardrail port-forward svc/yardrail-frontend 8080:80

Open http://localhost:8080 — you should see the Yardrail login page. You can also verify API access with the yr CLI:

yr auth login # authenticate against http://localhost:8080
yr config get # confirm the platform responds

3. Check the worker logs​

kubectl -n yardrail logs deployment/yardrail-worker --tail=20

A healthy worker logs its startup configuration and begins polling:

{"level":"INFO","msg":"yardrail-worker starting","temporal":"temporal:7233","executor":"kubernetes",...}

If the worker is crash-looping, common causes are:

  • Temporal server unreachable (--temporal-address)
  • S3 endpoint unreachable or bucket doesn't exist (--s3-endpoint, --s3-bucket)
  • Missing CSI driver name when --executor=kubernetes is set

4. Verify Kubernetes executor (if enabled)​

These checks apply when worker.executor: kubernetes.

RBAC is in place​

kubectl -n yardrail auth can-i create pods \
--as=system:serviceaccount:yardrail:yardrail-worker

Expected: yes

kubectl -n yardrail auth can-i create pods/attach \
--as=system:serviceaccount:yardrail:yardrail-worker

Expected: yes

CSI driver is running​

kubectl get csidrivers

Your S3 CSI driver should be listed (e.g., s3.csi.aws.amazon.com).

Test Pod creation​

To verify the worker can create and attach to Pods, trigger a workflow that uses a container tool. Watch for the CWL tool Pod:

kubectl -n yardrail get pods -l app.kubernetes.io/component=cwl-tool --watch

You should see a Pod appear, transition through Pending → Running → Completed, then get deleted (if Remove is set in the CWL tool spec).

Check S3 for artifacts​

After a workflow run, the output artifacts should be in S3 under the workflows/ prefix:

# AWS
aws s3 ls s3://yardrail-workflows/workflows/ --recursive | head -10

# MinIO
mc ls myminio/yardrail-workflows/workflows/ --recursive | head -10

Troubleshooting​

Worker can't create Pods​

creating pod: pods is forbidden: User "system:serviceaccount:yardrail:yardrail-worker"
cannot create resource "pods" in API group "" in the namespace "yardrail"

The RBAC Role or RoleBinding is missing. Verify worker.serviceAccount.create: true and worker.executor: kubernetes are both set in your Helm values, then upgrade the release.

Pod stuck in Pending​

kubectl -n yardrail describe pod <cwl-tool-pod-name>

Common causes:

  • CSI driver not installed: driver name s3.csi.aws.amazon.com not found — install the CSI driver per the setup guide.
  • Image pull failure: ErrImagePull — check that the image exists and workflowImagePullSecrets is configured for private registries.
  • Insufficient resources: the cluster has no nodes with available capacity.

SPDY attach fails​

creating SPDY executor: error upgrading connection

This usually means the Kubernetes API server doesn't support SPDY connections. Verify your cluster version is 1.27+ and that any network proxy between the worker and the API server allows connection upgrades.

Outputs missing from S3​

If the workflow succeeds but outputs aren't in S3, check:

  1. The CSI driver is correctly configured with the right bucket and credentials.
  2. The worker.s3.bucket value matches the bucket the CSI driver is configured for.
  3. The CSI driver's volumeAttributes include any required auth (endpoint, region, credentials).