Fix firecracker process cleanup and shutdown latency

Author: cursoragent

lib/hypervisor/firecracker/process.go

@@ -148,6 +148,10 @@ func (s *Starter) startProcess(_ context.Context, p *paths.Paths, version string
 	}
 
 	if err := waitForSocket(socketPath, socketWaitTimeout); err != nil {
+		if cmd.Process != nil {
+			_ = cmd.Process.Kill()
+			_ = cmd.Wait()
+		}
 		if data, readErr := os.ReadFile(vmmLogPath); readErr == nil && len(data) > 0 {
 			return 0, fmt.Errorf("%w; vmm.log: %s", err, string(data))
 		}

lib/instances/standby.go

@@ -190,8 +190,7 @@ func (m *manager) shutdownHypervisor(ctx context.Context, inst *Instance) error
 
 	// Wait for process to exit
 	if inst.HypervisorPID != nil {
-		if !WaitForProcessExit(*inst.HypervisorPID, 2*time.Second) {
-			log.WarnContext(ctx, "hypervisor did not exit gracefully in time, force killing process", "instance_id", inst.Id, "pid", *inst.HypervisorPID)
+		forceKill := func() error {
 			if err := syscall.Kill(*inst.HypervisorPID, syscall.SIGKILL); err != nil && err != syscall.ESRCH {
 				return fmt.Errorf("force kill hypervisor pid %d: %w", *inst.HypervisorPID, err)
 			}
@@ -202,6 +201,19 @@ func (m *manager) shutdownHypervisor(ctx context.Context, inst *Instance) error
 					return fmt.Errorf("hypervisor pid %d did not exit after SIGKILL", *inst.HypervisorPID)
 				}
 			}
+			return nil
+		}
+
+		if shutdownErr == hypervisor.ErrNotSupported {
+			log.DebugContext(ctx, "hypervisor does not support graceful shutdown, force killing process", "instance_id", inst.Id, "pid", *inst.HypervisorPID)
+			if err := forceKill(); err != nil {
+				return err
+			}
+		} else if !WaitForProcessExit(*inst.HypervisorPID, 2*time.Second) {
+			log.WarnContext(ctx, "hypervisor did not exit gracefully in time, force killing process", "instance_id", inst.Id, "pid", *inst.HypervisorPID)
+			if err := forceKill(); err != nil {
+				return err
+			}
 		} else {
 			log.DebugContext(ctx, "hypervisor shutdown gracefully", "instance_id", inst.Id, "pid", *inst.HypervisorPID)
 		}