[cri] add sandbox and container latency metrics

These are simple metrics that allow users to view more fine grained metrics on
internal operations.

Signed-off-by: Michael Crosby <michael@thepasture.io>

pkg/cri/server/container_start.go

@@ -40,11 +40,17 @@ import (
 
 // StartContainer starts the container.
 func (c *criService) StartContainer(ctx context.Context, r *runtime.StartContainerRequest) (retRes *runtime.StartContainerResponse, retErr error) {
+	start := time.Now()
 	cntr, err := c.containerStore.Get(r.GetContainerId())
 	if err != nil {
 		return nil, errors.Wrapf(err, "an error occurred when try to find container %q", r.GetContainerId())
 	}
 
+	info, err := cntr.Container.Info(ctx)
+	if err != nil {
+		return nil, errors.Wrap(err, "get container info")
+	}
+
 	id := cntr.ID
 	meta := cntr.Metadata
 	container := cntr.Container
@@ -162,6 +168,8 @@ func (c *criService) StartContainer(ctx context.Context, r *runtime.StartContain
 	// It handles the TaskExit event and update container state after this.
 	c.eventMonitor.startContainerExitMonitor(context.Background(), id, task.Pid(), exitCh)
 
+	containerStartTimer.WithValues(info.Runtime.Name).UpdateSince(start)
+
 	return &runtime.StartContainerResponse{}, nil
 }