feat: add resilient container restart controls

internal/updateagent/docker_test.go

@@ -1,13 +1,26 @@
 package updateagent
 
 import (
+	"context"
 	"encoding/json"
+	"errors"
+	"io"
+	"net/http"
 	"strconv"
+	"strings"
+	"sync/atomic"
 	"testing"
+	"time"
 
 	"github.com/root-kdw12/hetzner-api-dyndns-records-webinterface/internal/buildinfo"
 )
 
+type roundTripFunc func(*http.Request) (*http.Response, error)
+
+func (f roundTripFunc) RoundTrip(request *http.Request) (*http.Response, error) {
+	return f(request)
+}
+
 func TestValidateUpdateCompatibilityAllowsDirectMajorJump(t *testing.T) {
 	t.Parallel()
 	current := compatibleImage("v1.4.0", "v1.4.0")
@@ -17,6 +30,40 @@ func TestValidateUpdateCompatibilityAllowsDirectMajorJump(t *testing.T) {
 	}
 }
 
+func TestWaitForDockerAPIRetriesBootTransient(t *testing.T) {
+	t.Parallel()
+	var calls atomic.Int32
+	err := waitForDockerAPI(context.Background(), 100*time.Millisecond, time.Millisecond, func(context.Context) error {
+		if calls.Add(1) < 3 {
+			return errors.New("Docker is still starting")
+		}
+		return nil
+	})
+	if err != nil || calls.Load() != 3 {
+		t.Fatalf("Docker boot transient was not retried: calls=%d err=%v", calls.Load(), err)
+	}
+}
+
+func TestWaitCurrentHealthyAllowsStartingButRejectsUnhealthy(t *testing.T) {
+	t.Parallel()
+	var calls atomic.Int32
+	client := dockerTestClient(func() string {
+		if calls.Add(1) == 1 {
+			return "starting"
+		}
+		return "healthy"
+	})
+	inspected, err := client.waitCurrentHealthy(context.Background(), "container-id", "sha256:current", 100*time.Millisecond, time.Millisecond)
+	if err != nil || inspected.State.Health == nil || inspected.State.Health.Status != "healthy" {
+		t.Fatalf("starting container did not become eligible: inspect=%#v err=%v", inspected, err)
+	}
+
+	unhealthy := dockerTestClient(func() string { return "unhealthy" })
+	if _, err := unhealthy.waitCurrentHealthy(context.Background(), "container-id", "sha256:current", 100*time.Millisecond, time.Millisecond); err == nil || !strings.Contains(err.Error(), "unhealthy") {
+		t.Fatalf("unhealthy container was not rejected immediately: %v", err)
+	}
+}
+
 func TestResolveCurrentVersionRequiresRuntimeAndImageMetadataToAgree(t *testing.T) {
 	t.Parallel()
 	config := json.RawMessage(`{"Image":"registry.example/app:v1.4.0","Labels":{"org.opencontainers.image.version":"v1.4.0"}}`)
@@ -203,3 +250,11 @@ func contains(values []string, expected string) bool {
 	}
 	return false
 }
+
+func dockerTestClient(health func() string) *dockerClient {
+	transport := roundTripFunc(func(*http.Request) (*http.Response, error) {
+		body := `{"Id":"container-id","Image":"sha256:current","Config":{"Labels":{}},"State":{"Running":true,"ExitCode":0,"Health":{"Status":"` + health() + `"}}}`
+		return &http.Response{StatusCode: http.StatusOK, Header: make(http.Header), Body: io.NopCloser(strings.NewReader(body))}, nil
+	})
+	return &dockerClient{client: &http.Client{Transport: transport}, apiPrefix: "/v1.0"}
+}