milvus/tests/integration/util_insert.go
wei liu b54b9e0dfa
enhance: add channel balanced primary key generator for integration tests (#46431)
issue: #46352

- Add GenerateChannelBalancedPrimaryKeys function supporting Int64 and
VarChar PK types with even distribution across channels
- Add GenerateBalancedInt64PKs and GenerateBalancedVarCharPKs helper
functions using murmur3 and crc32 hash algorithms respectively
- Add PrimaryKeyConfig struct to configure PK generation in
InsertAndFlush
- Update InsertAndFlush to use channel balanced PKs instead of random
hash keys for better test coverage
- Add comprehensive unit tests including end-to-end validation with
typeutil.HashPK2Channels to verify exact channel distribution

<!-- This is an auto-generated comment: release notes by coderabbit.ai
-->
## Summary by CodeRabbit

* **Tests**
* Enhanced integration tests with configurable primary-key sequencing
across insert batches.
* Added utilities to generate channel-balanced primary keys for integer
and string types.
* Expanded test coverage validating balanced distribution, uniqueness,
continuation, and large-scale behavior across multiple channels.

<sub>✏️ Tip: You can customize this high-level summary in your review
settings.</sub>
<!-- end of auto-generated comment: release notes by coderabbit.ai -->

---------

Signed-off-by: wayblink <anyang.wang@zilliz.com>
Signed-off-by: Wei Liu <wei.liu@zilliz.com>
2025-12-26 14:49:19 +08:00

440 lines
13 KiB
Go

// Licensed to the LF AI & Data foundation under one
// or more contributor license agreements. See the NOTICE file
// distributed with this work for additional information
// regarding copyright ownership. The ASF licenses this file
// to you under the Apache License, Version 2.0 (the
// "License"); you may not use this file except in compliance
// with the License. You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
package integration
import (
"context"
"encoding/binary"
"fmt"
"hash/crc32"
"time"
"github.com/spaolacci/murmur3"
"github.com/milvus-io/milvus-proto/go-api/v2/milvuspb"
"github.com/milvus-io/milvus-proto/go-api/v2/schemapb"
"github.com/milvus-io/milvus/pkg/v2/util/testutils"
)
func (s *MiniClusterSuite) WaitForFlush(ctx context.Context, segIDs []int64, flushTs uint64, dbName, collectionName string) {
flushed := func() bool {
resp, err := s.Cluster.MilvusClient.GetFlushState(ctx, &milvuspb.GetFlushStateRequest{
SegmentIDs: segIDs,
FlushTs: flushTs,
DbName: dbName,
CollectionName: collectionName,
})
if err != nil {
return false
}
return resp.GetFlushed()
}
for !flushed() {
select {
case <-ctx.Done():
s.FailNow("failed to wait for flush until ctx done")
return
default:
time.Sleep(500 * time.Millisecond)
}
}
}
func NewInt64FieldData(fieldName string, numRows int) *schemapb.FieldData {
return &schemapb.FieldData{
Type: schemapb.DataType_Int64,
FieldName: fieldName,
Field: &schemapb.FieldData_Scalars{
Scalars: &schemapb.ScalarField{
Data: &schemapb.ScalarField_LongData{
LongData: &schemapb.LongArray{
Data: GenerateInt64Array(numRows, 0),
},
},
},
},
}
}
func NewInt64FieldDataWithStart(fieldName string, numRows int, start int64) *schemapb.FieldData {
return &schemapb.FieldData{
Type: schemapb.DataType_Int64,
FieldName: fieldName,
Field: &schemapb.FieldData_Scalars{
Scalars: &schemapb.ScalarField{
Data: &schemapb.ScalarField_LongData{
LongData: &schemapb.LongArray{
Data: GenerateInt64Array(numRows, start),
},
},
},
},
}
}
func NewInt64FieldDataNullableWithStart(fieldName string, numRows, start int) *schemapb.FieldData {
validData, num := GenerateBoolArray(numRows)
return &schemapb.FieldData{
Type: schemapb.DataType_Int64,
FieldName: fieldName,
Field: &schemapb.FieldData_Scalars{
Scalars: &schemapb.ScalarField{
Data: &schemapb.ScalarField_LongData{
LongData: &schemapb.LongArray{
Data: GenerateInt64Array(num, int64(start)),
},
},
},
},
ValidData: validData,
}
}
func NewInt64SameFieldData(fieldName string, numRows int, value int64) *schemapb.FieldData {
return &schemapb.FieldData{
Type: schemapb.DataType_Int64,
FieldName: fieldName,
Field: &schemapb.FieldData_Scalars{
Scalars: &schemapb.ScalarField{
Data: &schemapb.ScalarField_LongData{
LongData: &schemapb.LongArray{
Data: GenerateSameInt64Array(numRows, value),
},
},
},
},
}
}
func NewVarCharSameFieldData(fieldName string, numRows int, value string) *schemapb.FieldData {
return &schemapb.FieldData{
Type: schemapb.DataType_String,
FieldName: fieldName,
Field: &schemapb.FieldData_Scalars{
Scalars: &schemapb.ScalarField{
Data: &schemapb.ScalarField_StringData{
StringData: &schemapb.StringArray{
Data: GenerateSameStringArray(numRows, value),
},
},
},
},
}
}
func NewVarCharFieldData(fieldName string, numRows int, nullable bool) *schemapb.FieldData {
numValid := numRows
if nullable {
numValid = numRows / 2
}
return &schemapb.FieldData{
Type: schemapb.DataType_String,
FieldName: fieldName,
Field: &schemapb.FieldData_Scalars{
Scalars: &schemapb.ScalarField{
Data: &schemapb.ScalarField_StringData{
StringData: &schemapb.StringArray{
Data: testutils.GenerateStringArray(numValid),
// Data: testutils.GenerateStringArray(numRows),
},
},
},
},
ValidData: testutils.GenerateBoolArray(numRows),
}
}
func NewStringFieldData(fieldName string, numRows int) *schemapb.FieldData {
return testutils.NewStringFieldData(fieldName, numRows)
}
// note: unlike testutils's NewGeometryFieldData ,integration's NewGeometryFieldData generate wkt string bytes
func NewGeometryFieldData(fieldName string, numRows int) *schemapb.FieldData {
return testutils.NewGeometryFieldDataWktFormat(fieldName, numRows)
}
func NewFloatVectorFieldData(fieldName string, numRows, dim int) *schemapb.FieldData {
return testutils.NewFloatVectorFieldData(fieldName, numRows, dim)
}
func NewFloat16VectorFieldData(fieldName string, numRows, dim int) *schemapb.FieldData {
return testutils.NewFloat16VectorFieldData(fieldName, numRows, dim)
}
func NewBFloat16VectorFieldData(fieldName string, numRows, dim int) *schemapb.FieldData {
return testutils.NewBFloat16VectorFieldData(fieldName, numRows, dim)
}
func NewBinaryVectorFieldData(fieldName string, numRows, dim int) *schemapb.FieldData {
return testutils.NewBinaryVectorFieldData(fieldName, numRows, dim)
}
func NewSparseFloatVectorFieldData(fieldName string, numRows int) *schemapb.FieldData {
return testutils.NewSparseFloatVectorFieldData(fieldName, numRows)
}
func NewInt8VectorFieldData(fieldName string, numRows, dim int) *schemapb.FieldData {
return testutils.NewInt8VectorFieldData(fieldName, numRows, dim)
}
func NewStructArrayFieldData(schema *schemapb.StructArrayFieldSchema, fieldName string, numRow int, dim int) *schemapb.FieldData {
fieldData := &schemapb.FieldData{
Type: schemapb.DataType_ArrayOfStruct,
FieldName: fieldName,
Field: &schemapb.FieldData_StructArrays{
StructArrays: &schemapb.StructArrayField{
Fields: testutils.GenerateArrayOfStructArray(schema, numRow, dim),
},
},
}
return fieldData
}
func GenerateInt64Array(numRows int, start int64) []int64 {
ret := make([]int64, numRows)
for i := 0; i < numRows; i++ {
ret[i] = int64(i) + start
}
return ret
}
func GenerateSameInt64Array(numRows int, value int64) []int64 {
ret := make([]int64, numRows)
for i := 0; i < numRows; i++ {
ret[i] = value
}
return ret
}
func GenerateBoolArray(numRows int) ([]bool, int) {
var num int
ret := make([]bool, numRows)
for i := 0; i < numRows; i++ {
ret[i] = i%2 == 0
if ret[i] {
num++
}
}
return ret, num
}
func GenerateSameStringArray(numRows int, value string) []string {
ret := make([]string, numRows)
for i := 0; i < numRows; i++ {
ret[i] = value
}
return ret
}
func GenerateSparseFloatArray(numRows int) *schemapb.SparseFloatArray {
return testutils.GenerateSparseFloatVectors(numRows)
}
func GenerateHashKeys(numRows int) []uint32 {
return testutils.GenerateHashKeys(numRows)
}
// GenerateChannelBalancedPrimaryKeys generates primary keys that are evenly distributed across channels.
// It supports both Int64 and VarChar primary key types.
// For Int64: uses murmur3 hash (same as typeutil.Hash32Int64)
// For VarChar: uses crc32 hash (same as typeutil.HashString2Uint32)
// startPK specifies where to begin searching for PKs.
// Returns the FieldData and the next startPK for subsequent calls.
func GenerateChannelBalancedPrimaryKeys(fieldName string, fieldType schemapb.DataType, numRows int, numChannels int, startPK int64) (*schemapb.FieldData, int64) {
switch fieldType {
case schemapb.DataType_Int64:
pks, nextPK := GenerateBalancedInt64PKs(numRows, numChannels, startPK)
return &schemapb.FieldData{
Type: schemapb.DataType_Int64,
FieldName: fieldName,
Field: &schemapb.FieldData_Scalars{
Scalars: &schemapb.ScalarField{
Data: &schemapb.ScalarField_LongData{
LongData: &schemapb.LongArray{
Data: pks,
},
},
},
},
}, nextPK
case schemapb.DataType_VarChar, schemapb.DataType_String:
pks, nextIndex := GenerateBalancedVarCharPKs(numRows, numChannels, int(startPK))
return &schemapb.FieldData{
Type: schemapb.DataType_VarChar,
FieldName: fieldName,
Field: &schemapb.FieldData_Scalars{
Scalars: &schemapb.ScalarField{
Data: &schemapb.ScalarField_StringData{
StringData: &schemapb.StringArray{
Data: pks,
},
},
},
},
}, int64(nextIndex)
default:
panic(fmt.Sprintf("not supported primary key type: %s", fieldType))
}
}
// GenerateBalancedInt64PKs generates int64 primary keys that are evenly distributed across channels.
// This ensures each channel receives exactly numRows/numChannels items based on PK hash values.
// The function searches for PKs that hash to each channel to achieve exact distribution.
// startPK specifies where to begin searching for PKs.
// Returns the generated PKs and the next startPK for subsequent calls.
func GenerateBalancedInt64PKs(numRows int, numChannels int, startPK int64) ([]int64, int64) {
if numChannels <= 0 {
numChannels = 1
}
// Calculate how many items each channel should receive
baseCount := numRows / numChannels
remainder := numRows % numChannels
// Collect PKs for each channel
channelPKs := make([][]int64, numChannels)
targetCounts := make([]int, numChannels)
for ch := 0; ch < numChannels; ch++ {
targetCounts[ch] = baseCount
if ch < remainder {
targetCounts[ch]++
}
channelPKs[ch] = make([]int64, 0, targetCounts[ch])
}
// Search for PKs that hash to each channel
var lastPK int64
for pk := startPK; ; pk++ {
lastPK = pk
// Calculate which channel this PK would go to
hash := hashInt64ForChannel(pk)
ch := int(hash % uint32(numChannels))
if len(channelPKs[ch]) < targetCounts[ch] {
channelPKs[ch] = append(channelPKs[ch], pk)
// Check if all channels have enough PKs
done := true
for ch := 0; ch < numChannels; ch++ {
if len(channelPKs[ch]) < targetCounts[ch] {
done = false
break
}
}
if done {
break
}
}
}
// Combine all PKs
result := make([]int64, 0, numRows)
for ch := 0; ch < numChannels; ch++ {
result = append(result, channelPKs[ch]...)
}
return result, lastPK + 1
}
// hashInt64ForChannel computes the hash value for channel assignment.
// This mirrors the logic in typeutil.Hash32Int64 and HashPK2Channels.
func hashInt64ForChannel(v int64) uint32 {
// Must match the behavior of typeutil.Hash32Int64
// which uses common.Endian (binary.LittleEndian)
b := make([]byte, 8)
binary.LittleEndian.PutUint64(b, uint64(v))
// Use murmur3 hash (same as typeutil.Hash32Bytes)
h := murmur3.New32()
h.Write(b)
return h.Sum32() & 0x7fffffff
}
// GenerateBalancedVarCharPKs generates varchar primary keys that are evenly distributed across channels.
// This ensures each channel receives exactly numRows/numChannels items based on PK hash values.
// The function searches for PKs that hash to each channel to achieve exact distribution.
// startIndex specifies where to begin searching for PKs (used in "pk_<index>" format).
// Returns the generated PKs and the next startIndex for subsequent calls.
func GenerateBalancedVarCharPKs(numRows int, numChannels int, startIndex int) ([]string, int) {
if numChannels <= 0 {
numChannels = 1
}
// Calculate how many items each channel should receive
baseCount := numRows / numChannels
remainder := numRows % numChannels
// Collect PKs for each channel
channelPKs := make([][]string, numChannels)
targetCounts := make([]int, numChannels)
for ch := 0; ch < numChannels; ch++ {
targetCounts[ch] = baseCount
if ch < remainder {
targetCounts[ch]++
}
channelPKs[ch] = make([]string, 0, targetCounts[ch])
}
// Search for PKs that hash to each channel
var lastIndex int
for i := startIndex; ; i++ {
lastIndex = i
// Generate a unique string PK
pk := fmt.Sprintf("pk_%d", i)
// Calculate which channel this PK would go to
hash := hashVarCharForChannel(pk)
ch := int(hash % uint32(numChannels))
if len(channelPKs[ch]) < targetCounts[ch] {
channelPKs[ch] = append(channelPKs[ch], pk)
// Check if all channels have enough PKs
done := true
for ch := 0; ch < numChannels; ch++ {
if len(channelPKs[ch]) < targetCounts[ch] {
done = false
break
}
}
if done {
break
}
}
}
// Combine all PKs
result := make([]string, 0, numRows)
for ch := 0; ch < numChannels; ch++ {
result = append(result, channelPKs[ch]...)
}
return result, lastIndex + 1
}
// hashVarCharForChannel computes the hash value for channel assignment of varchar PKs.
// This mirrors the logic in typeutil.HashString2Uint32 and HashPK2Channels.
func hashVarCharForChannel(v string) uint32 {
// Must match the behavior of typeutil.HashString2Uint32
// which uses crc32.ChecksumIEEE with substring limit of 100 chars
subString := v
if len(v) > 100 {
subString = v[:100]
}
return crc32.ChecksumIEEE([]byte(subString))
}