shim: exit out of oom polling if unimplemented

This exits out of polling for OOM events if the getOOMEvent method is unimplemented. Signed-off-by: Alex Price <[email protected]>
kata-containers · May 7, 2020 · 86f5810 · 86f5810
1 parent b4833a4
commit 86f5810
Show file tree

Hide file tree

Showing 5 changed files with 31 additions and 3 deletions.
diff --git a/containerd-shim-v2/errors.go b/containerd-shim-v2/errors.go
@@ -60,3 +60,11 @@ func isNotFound(err error) bool {
 	return err == vc.ErrNoSuchContainer || err == syscall.ENOENT ||
 		strings.Contains(err.Error(), "not found") || strings.Contains(err.Error(), "not exist")
 }
+
+func isGRPCErrorCode(code codes.Code, err error) bool {
+	s, ok := status.FromError(err)
+	if !ok {
+		return false
+	}
+	return s != nil && s.Code() == code
+}
diff --git a/containerd-shim-v2/errors_test.go b/containerd-shim-v2/errors_test.go
@@ -6,11 +6,14 @@
 package containerdshim
 
 import (
+	"errors"
 	"syscall"
 	"testing"
 
 	vc "github.com/kata-containers/runtime/virtcontainers/pkg/types"
 	"github.com/stretchr/testify/assert"
+	"google.golang.org/grpc/codes"
+	"google.golang.org/grpc/status"
 )
 
 func TestToGRPC(t *testing.T) {
@@ -27,3 +30,10 @@ func TestToGRPC(t *testing.T) {
 		assert.True(isGRPCError(err))
 	}
 }
+
+func TestIsGRPCErrorCode(t *testing.T) {
+	assert := assert.New(t)
+
+	assert.True(isGRPCErrorCode(codes.Unimplemented, status.New(codes.Unimplemented, "foobar").Err()))
+	assert.False(isGRPCErrorCode(codes.Unimplemented, errors.New("foobar")))
+}
diff --git a/containerd-shim-v2/start.go b/containerd-shim-v2/start.go
@@ -37,8 +37,9 @@ func startContainer(ctx context.Context, s *service, c *container) error {
 		}
 		go watchSandbox(s)
 
-		// Start watching for oom events
-		go watchOOMEvents(ctx, s)
+		// We don't rely on the context passed to startContainer as it can be cancelled after
+		// this rpc call.
+		go watchOOMEvents(s.ctx, s)
 	} else {
 		_, err := s.sandbox.StartContainer(c.id)
 		if err != nil {

diff --git a/containerd-shim-v2/start_test.go b/containerd-shim-v2/start_test.go
@@ -45,6 +45,7 @@ func TestStartStartSandboxSuccess(t *testing.T) {
 		id:         testSandboxID,
 		sandbox:    sandbox,
 		containers: make(map[string]*container),
+		ctx:        context.Background(),
 	}
 
 	reqCreate := &taskAPI.CreateTaskRequest{
@@ -93,6 +94,7 @@ func TestStartMissingAnnotation(t *testing.T) {
 		id:         testSandboxID,
 		sandbox:    sandbox,
 		containers: make(map[string]*container),
+		ctx:        context.Background(),
 	}
 
 	reqCreate := &taskAPI.CreateTaskRequest{
@@ -159,6 +161,7 @@ func TestStartStartContainerSucess(t *testing.T) {
 		id:         testSandboxID,
 		sandbox:    sandbox,
 		containers: make(map[string]*container),
+		ctx:        context.Background(),
 	}
 
 	reqCreate := &taskAPI.CreateTaskRequest{

diff --git a/containerd-shim-v2/wait.go b/containerd-shim-v2/wait.go
@@ -14,6 +14,7 @@ import (
 	"github.com/containerd/containerd/api/types/task"
 	"github.com/containerd/containerd/mount"
 	"github.com/sirupsen/logrus"
+	"google.golang.org/grpc/codes"
 )
 
 func wait(s *service, c *container, execID string) (int32, error) {
@@ -141,7 +142,12 @@ func watchOOMEvents(ctx context.Context, s *service) {
 		default:
 			containerID, err := s.sandbox.GetOOMEvent()
 			if err != nil {
-				logrus.WithError(err).Warn("failed to get oom event from sandbox")
+				logrus.WithField("sandbox", s.sandbox.ID()).WithError(err).Warn("failed to get OOM event from sandbox")
+				// If the GetOOMEvent call is not implemented, then the agent is most likely an older version,
+				// stop attempting to get OOM events.
+				if isGRPCErrorCode(codes.Unimplemented, err) {
+					return
+				}
 				continue
 			}