Fix e2e runner missing docker CLI on Debian trixie

.github/workflows/e2e.yml

@@ -0,0 +1,29 @@
+name: e2e
+
+on:
+  push:
+    tags:
+      - v*
+    branches:
+      - master
+  pull_request:
+
+permissions:
+  contents: read
+
+jobs:
+  e2e:
+    name: e2e
+    runs-on: ubuntu-latest
+    steps:
+      - name: Check out code into the Go module directory
+        uses: actions/checkout@v3
+        with:
+          fetch-depth: 0
+          submodules: 'recursive'
+
+      - name: Run e2e tests in docker compose runner
+        run: |
+          docker compose -f e2e/docker-compose.yml up -d docker-sshd-e2e-target
+          trap "docker compose -f e2e/docker-compose.yml down || true" EXIT
+          docker compose -f e2e/docker-compose.yml run --rm e2e-runner

.github/workflows/tests.yml

@@ -30,25 +30,3 @@ jobs:
 
       - name: Run unit tests
         run: go test ./...
-
-  e2e:
-    name: e2e
-    runs-on: ubuntu-latest
-    needs: ut
-    steps:
-      - name: Check out code into the Go module directory
-        uses: actions/checkout@v3
-        with:
-          fetch-depth: 0
-          submodules: 'recursive'
-
-      - name: Set up Go 1.x
-        uses: actions/setup-go@v4
-        with:
-          go-version: '1.26.x'
-          cache: true
-
-      - name: Run e2e smoke checks
-        run: |
-          go run ./cmd/docker-sshd --help > /dev/null
-          go run ./cmd/kube-sshd --help > /dev/null

e2e/Dockerfile.testrunner

@@ -0,0 +1,6 @@
+FROM golang:1.26
+
+RUN apt-get update && apt-get install -y --no-install-recommends openssh-client curl ca-certificates docker-cli && rm -rf /var/lib/apt/lists/* \
+    && curl -fsSL https://dl.k8s.io/release/v1.34.1/bin/linux/amd64/kubectl -o /usr/local/bin/kubectl \
+    && GOBIN=/usr/local/bin go install sigs.k8s.io/kind@v0.30.0 \
+    && chmod +x /usr/local/bin/kubectl

e2e/docker-compose.yml

@@ -0,0 +1,40 @@
+services:
+  docker-sshd-e2e-target:
+    image: alpine:3.20
+    container_name: docker-sshd-e2e-test
+    command: ["sleep", "300"]
+  e2e-runner:
+    build:
+      context: ..
+      dockerfile: e2e/Dockerfile.testrunner
+    depends_on:
+      - docker-sshd-e2e-target
+    privileged: true
+    working_dir: /src
+    environment:
+      - KUBECONFIG=/root/.kube/config
+    volumes:
+      - ..:/src
+      - /var/run/docker.sock:/var/run/docker.sock
+      - ${KUBECONFIG:-/tmp/invalid-kubeconfig-path}:/root/.kube/config:ro
+    command:
+      - /bin/sh
+      - -ec
+      - |
+        export GOFLAGS=-buildvcs=false
+        mkdir -p /tmp/bin
+        go build -o /tmp/bin/docker-sshd ./cmd/docker-sshd
+        go build -o /tmp/bin/kube-sshd ./cmd/kube-sshd
+        if [ "$${KUBE_SSHD_E2E:-1}" = "1" ]; then
+          kind get clusters | grep -qx docker-sshd-e2e || kind create cluster --name docker-sshd-e2e
+          kind export kubeconfig --name docker-sshd-e2e --internal
+          kubectl wait --for=condition=Ready pod -n kube-system --all --timeout=2m
+        fi
+        DOCKER_SSHD_E2E=$${DOCKER_SSHD_E2E:-1} \
+        KUBE_SSHD_E2E=$${KUBE_SSHD_E2E:-1} \
+        DOCKER_SSHD_BIN=/tmp/bin/docker-sshd \
+        KUBE_SSHD_BIN=/tmp/bin/kube-sshd \
+        go test ./e2e -v
+        if [ "$${KUBE_SSHD_E2E:-1}" = "1" ]; then
+          kind delete cluster --name docker-sshd-e2e || true
+        fi

e2e/e2e_test.go

@@ -0,0 +1,134 @@
+package e2e
+
+import (
+	"os"
+	"os/exec"
+	"strings"
+	"testing"
+	"time"
+)
+
+func TestDockerSSHD(t *testing.T) {
+	if os.Getenv("DOCKER_SSHD_E2E") != "1" {
+		t.Skip("set DOCKER_SSHD_E2E=1 to run")
+	}
+
+	bin := os.Getenv("DOCKER_SSHD_BIN")
+	if bin == "" {
+		t.Fatal("DOCKER_SSHD_BIN is required")
+	}
+
+	container := os.Getenv("DOCKER_E2E_CONTAINER")
+	if container == "" {
+		container = "docker-sshd-e2e-test"
+	}
+
+	key := t.TempDir() + "/docker-sshd-e2e-key"
+	mustRun(t, "ssh-keygen", "-t", "ed25519", "-N", "", "-f", key)
+
+	cmd := exec.Command(bin, "--address", "127.0.0.1", "--port", "2232", "--server-key", key)
+	cmd.Stdout = os.Stdout
+	cmd.Stderr = os.Stderr
+	if err := cmd.Start(); err != nil {
+		t.Fatalf("start docker-sshd: %v", err)
+	}
+	defer func() {
+		_ = cmd.Process.Kill()
+		_, _ = cmd.Process.Wait()
+	}()
+
+	out := retrySSH(t, "2232", container+"@127.0.0.1")
+	if !hasOKLine(out) {
+		t.Fatalf("expected ok, got %q", out)
+	}
+}
+
+func TestKubeSSHD(t *testing.T) {
+	if os.Getenv("KUBE_SSHD_E2E") != "1" {
+		t.Skip("set KUBE_SSHD_E2E=1 to run")
+	}
+
+	bin := os.Getenv("KUBE_SSHD_BIN")
+	if bin == "" {
+		t.Fatal("KUBE_SSHD_BIN is required")
+	}
+
+	pod := "kube-sshd-e2e-" + strings.ToLower(time.Now().Format("150405.000000000"))
+	pod = strings.ReplaceAll(pod, ".", "-")
+	mustRun(t, "kubectl", "run", pod, "--image=busybox:1.36", "--restart=Never", "--command", "--", "sleep", "300")
+	defer func() {
+		if out, err := run(t, "kubectl", "delete", "pod", pod, "--ignore-not-found=true", "--wait=false"); err != nil {
+			t.Logf("cleanup pod %s failed: %v\n%s", pod, err, out)
+		}
+	}()
+	mustRun(t, "kubectl", "wait", "--for=condition=Ready", "pod/"+pod, "--timeout=120s")
+
+	key := t.TempDir() + "/kube-sshd-e2e-key"
+	mustRun(t, "ssh-keygen", "-t", "ed25519", "-N", "", "-f", key)
+
+	cmd := exec.Command(bin, "--address", "127.0.0.1", "--port", "2233", "--server-key", key, "--namespace", "default")
+	cmd.Stdout = os.Stdout
+	cmd.Stderr = os.Stderr
+	if err := cmd.Start(); err != nil {
+		t.Fatalf("start kube-sshd: %v", err)
+	}
+	defer func() {
+		_ = cmd.Process.Kill()
+		_, _ = cmd.Process.Wait()
+	}()
+
+	out := retrySSH(t, "2233", pod+"@127.0.0.1")
+	if !hasOKLine(out) {
+		t.Fatalf("expected ok, got %q", out)
+	}
+}
+
+func hasOKLine(out string) bool {
+	for _, line := range strings.Split(out, "\n") {
+		if strings.TrimSpace(line) == "ok" {
+			return true
+		}
+	}
+	return false
+}
+
+func retrySSH(t *testing.T, port, userHost string) string {
+	t.Helper()
+
+	var out string
+	var err error
+	for range 30 {
+		out, err = run(t,
+			"ssh",
+			"-o", "StrictHostKeyChecking=no",
+			"-o", "UserKnownHostsFile=/dev/null",
+			"-o", "PreferredAuthentications=none",
+			"-p", port,
+			userHost,
+			"echo", "ok",
+		)
+		if err == nil {
+			return out
+		}
+		time.Sleep(time.Second)
+	}
+
+	t.Fatalf("ssh did not succeed: %v", err)
+	return ""
+}
+
+func mustRun(t *testing.T, name string, args ...string) string {
+	t.Helper()
+	out, err := run(t, name, args...)
+	if err != nil {
+		t.Fatalf("%s %v failed: %v\n%s", name, args, err, out)
+	}
+	return out
+}
+
+func run(t *testing.T, name string, args ...string) (string, error) {
+	t.Helper()
+	cmd := exec.Command(name, args...)
+	out, err := cmd.CombinedOutput()
+	return string(out), err
+}