Skaffold: Skaffold does not do any port-forwarding

[soullivaneuh]

Expected behavior

As the documentation says, a port-forwarding should be done base on my exposed port container.

Actual behavior

It does not a thing about that:

$ skaffold dev
Generating Tag for [localhost:32000/sweetforeign_web]...
Starting build...
Building [localhost:32000/sweetforeign_web]...
Sending build context to Docker daemon  3.072kB
Step 1/2 : FROM nginx:stable-alpine
 ---> 66952fd0a8ef
Step 2/2 : ADD index.html /usr/share/nginx/html
 ---> Using cache
 ---> f1cb0e1ba59c
Successfully built f1cb0e1ba59c
Successfully tagged localhost:32000/sweetforeign_web:latest
The push refers to repository [localhost:32000/sweetforeign_web]
5245eb10a6da: Preparing
129ba078f157: Preparing
8c8f1eccd524: Preparing
68442845474f: Preparing
503e53e365f3: Preparing
503e53e365f3: Layer already exists
68442845474f: Layer already exists
129ba078f157: Layer already exists
5245eb10a6da: Layer already exists
8c8f1eccd524: Layer already exists
latest: digest: sha256:47dc6256b50971da1e7e78154a919508bf04c9db0bcb9f1622dd71ee62eb1495 size: 1360
Build complete in 134.550577ms
Starting test...
Test complete in 4.058µs
Starting deploy...
Error: release: "sweetforeign" not found
Helm release sweetforeign not installed. Installing...
No requirements found in ./charts.
NAME:   sweetforeign
LAST DEPLOYED: Sat Feb 16 01:15:12 2019
NAMESPACE: default
STATUS: DEPLOYED

RESOURCES:
==> v1/Pod(related)
NAME                          READY  STATUS             RESTARTS  AGE
sweetforeign-cd89f9cf4-9fjc5  0/1    ContainerCreating  0         0s
sweetforeign-cd89f9cf4-vzz88  0/1    Terminating        0         17s

==> v1/Service
NAME          TYPE       CLUSTER-IP     EXTERNAL-IP  PORT(S)  AGE
sweetforeign  ClusterIP  10.152.183.97  <none>       80/TCP   0s

==> v1/Deployment
NAME          DESIRED  CURRENT  UP-TO-DATE  AVAILABLE  AGE
sweetforeign  1        1        1           0          0s


NOTES:
1. Get the application URL by running these commands:
  export POD_NAME=$(kubectl get pods --namespace default -l "app.kubernetes.io/name=sweetforeign,app.kubernetes.io/instance=sweetforeign" -o jsonpath="{.items[0].metadata.name}")
  echo "Visit http://127.0.0.1:8080 to use your application"
  kubectl port-forward $POD_NAME 8080:80

WARN[0001] error adding label to runtime object: patching resource /sweetforeign-test-connection: pods "sweetforeign-test-connection" not found 
Deploy complete in 1.35663968s
Watching for changes every 1s...

But it works with your helm example.

Information

• Skaffold version: 0.23.0
• Operating system: Linux sweethome 4.18.0-15-generic #16-Ubuntu SMP Thu Feb 7 10:56:39 UTC 2019 x86_64 x86_64 x86_64 GNU/Linux
• Contents of skaffold.yaml: See reproducing steps.

Steps to reproduce the behavior

1. Clone this branch: https://github.com/Soullivaneuh/skaffold-issues/tree/issue-1663
2. Run skaffold dev

[Pozo]

I confirm this. When I rolled back to v0.22.0 it seems to work.

[soullivaneuh]

@Pozo Did you test it on my issue branch? I just built v0.22.0 and it does not change a thing. :thinking:

[priyawadhwa]

@Soullivaneuh I checked out your branch and replaced the image name with my GCR project. Looks like this is happening because skaffold builds the image as gcr.io/priya-wadhwa/sweetforeign_web:latest@sha256:006f9aa59e6f68781481559cccab095f55750ee44d4931f13470cf185d0038f4, but helm deploys as gcr.io/priya-wadhwa/sweetforeign_web:latest, so our pod selector thinks the images are different and doesn't port forward.

We could potentially solve this by making the image selector less strict, or by adding a port-forward section to the skaffold.yaml to allow users to specify which containers to forward (this would be #1311)

[swoldanski]

@Soullivaneuh It is not a bug at all. I fall in the same trap.

Take a time to look at Helm example https://github.com/GoogleContainerTools/skaffold/tree/master/examples/helm-deployment

To make it work you have to change your deployment.yaml and values.yaml as well:

containers:
        - name: {{ .Chart.Name }}
          image: "{{ .Values.image }}"
          imagePullPolicy: {{ .Values.pullPolicy }}

After this change, your Chart will use proper image build by skaffold for deployment and port forwarding should work as expected.

[gbates101]

@swoldanski I'm experiencing the same behavior without using helm. My app port forwards just fine from 0.22.0-, but as soon as I upgrade to 0.23.0+ it stops. I would expect skaffold to auto port-forward with the following yaml:

apiVersion: extensions/v1beta1
kind: Deployment
metadata:
  name: app
  labels:
    app: app
spec:
  replicas: 1
  minReadySeconds: 10
  template:
    metadata:
      name: app-pod
      labels:
        app: app
    spec:
      containers:
      - name: app
        image: company/app:dev
        imagePullPolicy: Never
        ports:
        - name: http
          containerPort: 8000

Again, skaffold auto port-forwards fine with this yaml in <= 0.22.0, but not > 0.22.0.

[tjerkw]

I can confirm this, port forwarding is broken in skaffold version v0.25.0

[tjerkw]

This is an actual bug, or code works fine with v0.22.0 but port forwards breaks in v0.25.0

[masonlr]

Am experiencing this issue with v0.27.0. containerPort port forwarding is working with build artefacts defined in the skaffold yaml, but not with additional containers defined in the k8s manifests yaml.

[balopat]

With helm the latest version still has some issues (https://github.com/GoogleContainerTools/skaffold/issues/2379) but you can workaround it with explicitly configuring port-forwarding of deployments:

https://skaffold.dev/docs/how-tos/portforward/#user-defined-port-forwarding

[balopat]

I would close this in favor for #2379. Let me know if there is anything else here that justify reopening this issue.