enhance: Log error instead of panicking if load lock wait timeout (#39308)

Related to #39205
Previous PR #39206

This PR change wait timeout behavior to log error and return to avoid
making other collection read failure in only some collections have
deadlock

Signed-off-by: Congqi Xia <congqi.xia@zilliz.com>

internal/querynodev2/segments/segment.go

@@ -1176,10 +1176,10 @@ func (s *LocalSegment) WarmupChunkCache(ctx context.Context, fieldID int64, mmap
 		}).Await()
 	case "async":
 		task := func() (any, error) {
-			// bad implemtation, warmup is async at another goroutine and hold the rlock.
-			// the state transition of segment in segment loader will blocked.
-			// add a waiter to avoid it.
-			s.ptrLock.BlockUntilDataLoadedOrReleased()
+			// failed to wait for state update, return directly
+			if !s.ptrLock.BlockUntilDataLoadedOrReleased() {
+				return nil, nil
+			}
 			if s.PinIfNotReleased() != nil {
 				return nil, nil
 			}

internal/querynodev2/segments/state/load_state_lock.go

@@ -7,7 +7,9 @@ import (
 
 	"github.com/cockroachdb/errors"
 	"go.uber.org/atomic"
+	"go.uber.org/zap"
 
+	"github.com/milvus-io/milvus/pkg/log"
 	"github.com/milvus-io/milvus/pkg/util/paramtable"
 )
 
@@ -168,10 +170,12 @@ func (ls *LoadStateLock) StartReleaseAll() (g LoadStateLockGuard) {
 }
 
 // blockUntilDataLoadedOrReleased blocks until the segment is loaded or released.
-func (ls *LoadStateLock) BlockUntilDataLoadedOrReleased() {
+func (ls *LoadStateLock) BlockUntilDataLoadedOrReleased() bool {
+	var ok bool
 	ls.waitOrPanic(func(state loadStateEnum) bool {
 		return state == LoadStateDataLoaded || state == LoadStateReleased
-	}, noop)
+	}, func() { ok = true })
+	return ok
 }
 
 // waitUntilCanReleaseData waits until segment is release data able.
@@ -199,7 +203,7 @@ func (ls *LoadStateLock) waitOrPanic(ready func(state loadStateEnum) bool, then
 
 	select {
 	case <-time.After(maxWaitTime):
-		panic(fmt.Sprintf("max WLock wait time(%v) excceeded", maxWaitTime))
+		log.Error("load state lock wait timeout", zap.Duration("maxWaitTime", maxWaitTime))
 	case <-ch:
 	}
 }

internal/querynodev2/segments/state/load_state_lock_test.go

@@ -197,12 +197,15 @@ func TestWaitOrPanic(t *testing.T) {
 		defer paramtable.Get().Reset(paramtable.Get().CommonCfg.MaxWLockConditionalWaitTime.Key)
 
 		l := NewLoadStateLock(LoadStateDataLoaded)
+		executed := false
 
 		assert.NotPanics(t, func() {
 			l.waitOrPanic(func(state loadStateEnum) bool {
 				return state == LoadStateDataLoaded
-			}, noop)
+			}, func() { executed = true })
 		})
+
+		assert.True(t, executed)
 	})
 
 	t.Run("timeout_panic", func(t *testing.T) {
@@ -210,12 +213,14 @@ func TestWaitOrPanic(t *testing.T) {
 		defer paramtable.Get().Reset(paramtable.Get().CommonCfg.MaxWLockConditionalWaitTime.Key)
 
 		l := NewLoadStateLock(LoadStateOnlyMeta)
+		executed := false
 
-		assert.Panics(t, func() {
+		assert.NotPanics(t, func() {
 			l.waitOrPanic(func(state loadStateEnum) bool {
 				return state == LoadStateDataLoaded
 			}, noop)
 		})
+		assert.False(t, executed)
 	})
 }
 

pkg/util/hardware/gpu_mem_info_cuda.go

@@ -51,9 +51,11 @@ int getAllGPUMemoryInfo(GPUMemoryInfo** infos) {
 }
 */
 import "C"
+
 import (
-	"github.com/cockroachdb/errors"
 	"unsafe"
+
+	"github.com/cockroachdb/errors"
 )
 
 // GPUMemoryInfo represents a single GPU's memory information.