Fix spawn too many threads (milvus-io#26293)

- Low the thread pool cap
- Limit CGO calls concurrency

Signed-off-by: yah01 <[email protected]>

Author: yah01

.clang-tidy

@@ -18,7 +18,7 @@
 # The checks defined here will be run and will display by default as warnings.
 Checks: >
     -*, clang-diagnostic-*, -clang-diagnostic-error,
-    clang-analyzer-*, -clang-analyzer-alpha*,
+    clang-analyzer-*,
     google-*, -google-runtime-references, -google-readability-todo,
     modernize-*, -modernize-use-trailing-return-type, -modernize-use-nodiscard,
     performance-*,

configs/milvus.yaml

@@ -453,9 +453,9 @@ common:
   entityExpiration: -1 # Entity expiration in seconds, CAUTION make sure entityExpiration >= retentionDuration and -1 means never expire
   indexSliceSize: 16 # MB
   threadCoreCoefficient:
-    highPriority: 100 # This parameter specify how many times the number of threads is the number of cores in high priority thread pool
-    middlePriority: 50 # This parameter specify how many times the number of threads is the number of cores in middle priority thread pool
-    lowPriority: 10 # This parameter specify how many times the number of threads is the number of cores in low priority thread pool
+    highPriority: 10 # This parameter specify how many times the number of threads is the number of cores in high priority thread pool
+    middlePriority: 5 # This parameter specify how many times the number of threads is the number of cores in middle priority thread pool
+    lowPriority: 1 # This parameter specify how many times the number of threads is the number of cores in low priority thread pool
   DiskIndex:
     MaxDegree: 56
     SearchListSize: 100

internal/querynodev2/segments/segment_loader.go

@@ -301,6 +301,9 @@ func (loader *segmentLoader) notifyLoadFinish(segments ...*querypb.SegmentLoadIn
 func (loader *segmentLoader) requestResource(ctx context.Context, infos ...*querypb.SegmentLoadInfo) (LoadResource, int, error) {
 	resource := LoadResource{}
 
+	loader.mut.Lock()
+	defer loader.mut.Unlock()
+
 	memoryUsage := hardware.GetUsedMemoryCount()
 	totalMemory := hardware.GetMemoryCount()
 
@@ -310,9 +313,6 @@ func (loader *segmentLoader) requestResource(ctx context.Context, infos ...*quer
 	}
 	diskCap := paramtable.Get().QueryNodeCfg.DiskCapacityLimit.GetAsUint64()
 
-	loader.mut.Lock()
-	defer loader.mut.Unlock()
-
 	poolCap := runtime.NumCPU() * paramtable.Get().CommonCfg.HighPriorityThreadCoreCoefficient.GetAsInt()
 	if loader.committedResource.WorkNum >= poolCap {
 		return resource, 0, merr.WrapErrServiceRequestLimitExceeded(int32(poolCap))