chore: simplify cluster controller initialisation, prepare for reset

internal/machine/corroservice/docker.go

@@ -52,6 +52,20 @@ func (s *DockerService) Start(ctx context.Context) error {
 	return s.startNewContainer(ctx)
 }
 
+func (s *DockerService) Stop(ctx context.Context) error {
+	if err := s.Client.ContainerStop(ctx, s.Name, container.StopOptions{}); err != nil {
+		return fmt.Errorf("stop container %q: %w", s.Name, err)
+	}
+	slog.Debug("Corrosion Docker container stopped.", "name", s.Name)
+
+	if err := s.Client.ContainerRemove(ctx, s.Name, container.RemoveOptions{}); err != nil {
+		return fmt.Errorf("remove container %q: %w", s.Name, err)
+	}
+	slog.Debug("Corrosion Docker container removed.", "name", s.Name)
+
+	return nil
+}
+
 func (s *DockerService) Restart(ctx context.Context) error {
 	if err := s.Client.ContainerRestart(ctx, s.Name, container.StopOptions{}); err != nil {
 		return fmt.Errorf("restart container %q: %w", s.Name, err)

internal/machine/corroservice/service.go

@@ -4,6 +4,7 @@ import "context"
 
 type Service interface {
 	Start(ctx context.Context) error
+	Stop(ctx context.Context) error
 	Restart(ctx context.Context) error
 	Running() bool
 }

internal/machine/corroservice/subprocess.go

@@ -1,157 +0,0 @@
-package corroservice
-
-import (
-	"bufio"
-	"context"
-	"fmt"
-	"log/slog"
-	"os/exec"
-	"path/filepath"
-	"sync"
-	"syscall"
-	"time"
-)
-
-const (
-	DefaultCommand = "corrosion"
-	DefaultDataDir = "/var/lib/uncloud/corrosion"
-)
-
-// SubprocessService implements the Service interface by running the service as a subprocess.
-type SubprocessService struct {
-	Command string
-	DataDir string
-
-	cmd         *exec.Cmd
-	running     bool
-	mu          sync.Mutex
-	cancelWatch context.CancelFunc
-}
-
-func DefaultSubprocessService() *SubprocessService {
-	return &SubprocessService{
-		Command: DefaultCommand,
-		DataDir: DefaultDataDir,
-	}
-}
-
-// TODO: maybe stop the process if this ctx is cancelled.
-func (s *SubprocessService) Start(ctx context.Context) error {
-	s.mu.Lock()
-	defer s.mu.Unlock()
-
-	if s.running {
-		return nil
-	}
-	return s.startProcess(ctx)
-}
-
-func (s *SubprocessService) Restart(ctx context.Context) error {
-	s.mu.Lock()
-	defer s.mu.Unlock()
-
-	if s.running {
-		if err := s.stopProcess(); err != nil {
-			return fmt.Errorf("stop process: %w", err)
-		}
-	}
-	return s.startProcess(ctx)
-}
-
-func (s *SubprocessService) Running() bool {
-	s.mu.Lock()
-	defer s.mu.Unlock()
-	return s.running
-}
-
-func (s *SubprocessService) startProcess(ctx context.Context) error {
-	s.cmd = exec.Command(s.Command, "agent", "-c", filepath.Join(s.DataDir, "config.toml"))
-
-	// Redirect stdout and stderr to the logger.
-	stdout, err := s.cmd.StdoutPipe()
-	if err != nil {
-		return fmt.Errorf("create stdout pipe: %w", err)
-	}
-	stderr, err := s.cmd.StderrPipe()
-	if err != nil {
-		return fmt.Errorf("create stderr pipe: %w", err)
-	}
-
-	go func() {
-		scanner := bufio.NewScanner(stdout)
-		for scanner.Scan() {
-			slog.Info("[corrosion]: " + scanner.Text())
-		}
-		// TODO: remove
-		slog.Info("######## corrosion redirect go routine end ########")
-	}()
-
-	go func() {
-		scanner := bufio.NewScanner(stderr)
-		for scanner.Scan() {
-			slog.Error("[corrosion]: " + scanner.Text())
-		}
-	}()
-
-	if err = s.cmd.Start(); err != nil {
-		return fmt.Errorf("start process: %w", err)
-	}
-	s.running = true
-
-	// Watch for process exit to update running status.
-	go func() {
-		if err := s.cmd.Wait(); err != nil {
-			slog.Error("corrosion process exited with error.", "code", s.cmd.ProcessState.ExitCode(), "err", err)
-		}
-
-		s.mu.Lock()
-		s.running = false
-		s.mu.Unlock()
-	}()
-
-	// TODO: figure out the waiting process
-	// Wait for initialization
-	// timer := time.NewTimer(2 * time.Second)
-	// defer timer.Stop()
-
-	//select {
-	////case <-timer.C:
-	////	s.running = true
-	////	return nil
-	//case <-watchCtx.Done():
-	//	return fmt.Errorf("process failed to start")
-	//case <-ctx.Done():
-	//	s.stopProcess()
-	//	return ctx.Err()
-	//}
-	return nil
-}
-
-func (s *SubprocessService) stopProcess() error {
-	if s.cmd == nil || s.cmd.Process == nil {
-		return nil
-	}
-
-	if err := s.cmd.Process.Signal(syscall.SIGTERM); err != nil {
-		return fmt.Errorf("send SIGTERM: %w", err)
-	}
-
-	// Wait up to 5 seconds for graceful shutdown before killing the process.
-	done := make(chan error, 1)
-	go func() {
-		done <- s.cmd.Wait()
-	}()
-
-	select {
-	case <-time.After(5 * time.Second):
-		if err := s.cmd.Process.Kill(); err != nil {
-			return fmt.Errorf("kill process: %w", err)
-		}
-	case err := <-done:
-		if err != nil {
-			return fmt.Errorf("process exited with error: %w", err)
-		}
-	}
-
-	return nil
-}

internal/machine/corroservice/systemd.go

@@ -27,6 +27,15 @@ func (s *SystemdService) Start(ctx context.Context) error {
 	return s.startOrRestart(ctx, "start")
 }
 
+func (s *SystemdService) Stop(ctx context.Context) error {
+	if _, err := exec.Command("systemctl", "stop", s.Unit).Output(); err != nil {
+		return fmt.Errorf("systemctl stop %s: %w", s.Unit, err)
+	}
+	slog.Info("Corrosion systemd service stopped.", "unit", s.Unit)
+
+	return nil
+}
+
 func (s *SystemdService) Restart(ctx context.Context) error {
 	return s.startOrRestart(ctx, "restart")
 }