Merge pull request #2837 from jterry75/containerd_dump_stacks

Add dumpStacks support for containerd on Windows

cmd/containerd/command/main.go

@@ -24,6 +24,7 @@ import (
 	"os"
 	"os/signal"
 	"path/filepath"
+	"runtime"
 	"time"
 
 	"github.com/containerd/containerd/log"
@@ -232,3 +233,18 @@ func setLevel(context *cli.Context, config *srvconfig.Config) error {
 	}
 	return nil
 }
+
+func dumpStacks() {
+	var (
+		buf       []byte
+		stackSize int
+	)
+	bufferLen := 16384
+	for stackSize == len(buf) {
+		buf = make([]byte, bufferLen)
+		stackSize = runtime.Stack(buf, true)
+		bufferLen *= 2
+	}
+	buf = buf[:stackSize]
+	logrus.Infof("=== BEGIN goroutine stack dump ===\n%s\n=== END goroutine stack dump ===", buf)
+}

cmd/containerd/command/main_unix.go

@@ -21,11 +21,9 @@ package command
 import (
 	"context"
 	"os"
-	"runtime"
 
 	"github.com/containerd/containerd/log"
 	"github.com/containerd/containerd/services/server"
-	"github.com/sirupsen/logrus"
 	"golang.org/x/sys/unix"
 )
 
@@ -66,18 +64,3 @@ func handleSignals(ctx context.Context, signals chan os.Signal, serverC chan *se
 	}()
 	return done
 }
-
-func dumpStacks() {
-	var (
-		buf       []byte
-		stackSize int
-	)
-	bufferLen := 16384
-	for stackSize == len(buf) {
-		buf = make([]byte, bufferLen)
-		stackSize = runtime.Stack(buf, true)
-		bufferLen *= 2
-	}
-	buf = buf[:stackSize]
-	logrus.Infof("=== BEGIN goroutine stack dump ===\n%s\n=== END goroutine stack dump ===", buf)
-}

cmd/containerd/command/main_windows.go

@@ -18,12 +18,15 @@ package command
 
 import (
 	"context"
+	"fmt"
 	"os"
 	"path/filepath"
+	"unsafe"
 
+	winio "github.com/Microsoft/go-winio"
 	"github.com/containerd/containerd/log"
 	"github.com/containerd/containerd/services/server"
-
+	"github.com/sirupsen/logrus"
 	"golang.org/x/sys/windows"
 )
 
@@ -54,5 +57,35 @@ func handleSignals(ctx context.Context, signals chan os.Signal, serverC chan *se
 			}
 		}
 	}()
+	setupDumpStacks()
 	return done
 }
+
+func setupDumpStacks() {
+	// Windows does not support signals like *nix systems. So instead of
+	// trapping on SIGUSR1 to dump stacks, we wait on a Win32 event to be
+	// signaled. ACL'd to builtin administrators and local system
+	event := "Global\\containerd-daemon-" + fmt.Sprint(os.Getpid())
+	ev, _ := windows.UTF16PtrFromString(event)
+	sd, err := winio.SddlToSecurityDescriptor("D:P(A;;GA;;;BA)(A;;GA;;;SY)")
+	if err != nil {
+		logrus.Errorf("failed to get security descriptor for debug stackdump event %s: %s", event, err.Error())
+		return
+	}
+	var sa windows.SecurityAttributes
+	sa.Length = uint32(unsafe.Sizeof(sa))
+	sa.InheritHandle = 1
+	sa.SecurityDescriptor = uintptr(unsafe.Pointer(&sd[0]))
+	h, err := windows.CreateEvent(&sa, 0, 0, ev)
+	if h == 0 || err != nil {
+		logrus.Errorf("failed to create debug stackdump event %s: %s", event, err.Error())
+		return
+	}
+	go func() {
+		logrus.Debugf("Stackdump - waiting signal at %s", event)
+		for {
+			windows.WaitForSingleObject(h, windows.INFINITE)
+			dumpStacks()
+		}
+	}()
+}