milvus/internal/storagev2/packed/packed_reader_ffi.go
congqixia c01fd94a6a
enhance: integrate Storage V2 FFI interface for unified storage access (#45723)
Related #44956
This commit integrates the Storage V2 FFI (Foreign Function Interface)
interface throughout the Milvus codebase, enabling unified storage
access through the Loon FFI layer. This is a significant step towards
standardizing storage operations across different storage versions.

1. Configuration Support
- **configs/milvus.yaml**: Added `useLoonFFI` configuration flag under
`common.storage.file.splitByAvgSize` section
- Allows runtime toggle between traditional binlog readers and new
FFI-based manifest readers
  - Default: `false` (maintains backward compatibility)

2. Core FFI Infrastructure

Enhanced Utilities (internal/core/src/storage/loon_ffi/util.cpp/h)
- **ToCStorageConfig()**: Converts Go's `StorageConfig` to C's
`CStorageConfig` struct for FFI calls
- **GetManifest()**: Parses manifest JSON and retrieves latest column
groups using FFI
  - Accepts manifest path with `base_path` and `ver` fields
  - Calls `get_latest_column_groups()` FFI function
  - Returns column group information as string
  - Comprehensive error handling for JSON parsing and FFI errors

3. Dependency Updates
- **internal/core/thirdparty/milvus-storage/CMakeLists.txt**:
  - Updated milvus-storage version from `0883026` to `302143c`
  - Ensures compatibility with latest FFI interfaces

4. Data Coordinator Changes

All compaction task builders now include manifest path in segment
binlogs:

- **compaction_task_clustering.go**: Added `Manifest:
segInfo.GetManifestPath()` to segment binlogs
- **compaction_task_l0.go**: Added manifest path to both L0 segment
selection and compaction plan building
- **compaction_task_mix.go**: Added manifest path to mixed compaction
segment binlogs
- **meta.go**: Updated metadata completion logic:
- `completeClusterCompactionMutation()`: Set `ManifestPath` in new
segment info
- `completeMixCompactionMutation()`: Preserve manifest path in compacted
segments
- `completeSortCompactionMutation()`: Include manifest path in sorted
segments

5. Data Node Compactor Enhancements

All compactors updated to support dual-mode reading (binlog vs
manifest):

6. Flush & Sync Manager Updates

Pack Writer V2 (pack_writer_v2.go)
- **BulkPackWriterV2.Write()**: Extended return signature to include
`manifest string`
- Implementation:
  - Generate manifest path: `path.Join(pack.segmentID, "manifest.json")`
  - Write packed data using FFI-based writer
  - Return manifest path along with binlogs, deltas, and stats

Task Handling (task.go)
- Updated all sync task result handling to accommodate new manifest
return value
- Ensured backward compatibility for callers not using manifest

7. Go Storage Layer Integration

New Interfaces and Implementations
- **record_reader.go**: Interface for unified record reading across
storage versions
- **record_writer.go**: Interface for unified record writing across
storage versions
- **binlog_record_writer.go**: Concrete implementation for traditional
binlog-based writing

Enhanced Schema Support (schema.go, schema_test.go)
- Schema conversion utilities to support FFI-based storage operations
- Ensures proper Arrow schema mapping for V2 storage

Serialization Updates
- **serde.go, serde_events.go, serde_events_v2.go**: Updated to work
with new reader/writer interfaces
- Test files updated to validate dual-mode serialization

8. Storage V2 Packed Format

FFI Common (storagev2/packed/ffi_common.go)
- Common FFI utilities and type conversions for packed storage format

Packed Writer FFI (storagev2/packed/packed_writer_ffi.go)
- FFI-based implementation of packed writer
- Integrates with Loon storage layer for efficient columnar writes

Packed Reader FFI (storagev2/packed/packed_reader_ffi.go)
- Already existed, now complemented by writer implementation

9. Protocol Buffer Updates

data_coord.proto & datapb/data_coord.pb.go
- Added `manifest` field to compaction segment messages
- Enables passing manifest metadata through compaction pipeline

worker.proto & workerpb/worker.pb.go
- Added compaction parameter for `useLoonFFI` flag
- Allows workers to receive FFI configuration from coordinator

10. Parameter Configuration

component_param.go
- Added `UseLoonFFI` parameter to compaction configuration
- Reads from `common.storage.file.useLoonFFI` config path
- Default: `false` for safe rollout

11. Test Updates
- **clustering_compactor_storage_v2_test.go**: Updated signatures to
handle manifest return value
- **mix_compactor_storage_v2_test.go**: Updated test helpers for
manifest support
- **namespace_compactor_test.go**: Adjusted writer calls to expect
manifest
- **pack_writer_v2_test.go**: Validated manifest generation in pack
writing

This integration follows a **dual-mode approach**:
1. **Legacy Path**: Traditional binlog-based reading/writing (when
`useLoonFFI=false` or no manifest)
2. **FFI Path**: Manifest-based reading/writing through Loon FFI (when
`useLoonFFI=true` and manifest exists)

---------

Signed-off-by: Congqi Xia <congqi.xia@zilliz.com>
2025-11-24 19:57:07 +08:00

220 lines
7.3 KiB
Go

// Copyright 2023 Zilliz
//
// Licensed 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 packed
/*
#cgo pkg-config: milvus_core
#include <stdlib.h>
#include "storage/loon_ffi/ffi_reader_c.h"
#include "arrow/c/abi.h"
#include "arrow/c/helpers.h"
*/
import "C"
import (
"fmt"
"io"
"unsafe"
"github.com/apache/arrow/go/v17/arrow"
"github.com/apache/arrow/go/v17/arrow/cdata"
"go.uber.org/zap"
"github.com/milvus-io/milvus/pkg/v2/log"
"github.com/milvus-io/milvus/pkg/v2/proto/indexcgopb"
"github.com/milvus-io/milvus/pkg/v2/proto/indexpb"
)
func NewFFIPackedReader(manifest string, schema *arrow.Schema, neededColumns []string, bufferSize int64, storageConfig *indexpb.StorageConfig, storagePluginContext *indexcgopb.StoragePluginContext) (*FFIPackedReader, error) {
cManifest := C.CString(manifest)
defer C.free(unsafe.Pointer(cManifest))
var cas cdata.CArrowSchema
cdata.ExportArrowSchema(schema, &cas)
cSchema := (*C.struct_ArrowSchema)(unsafe.Pointer(&cas))
defer cdata.ReleaseCArrowSchema(&cas)
var cPackedReader C.CFFIPackedReader
var status C.CStatus
var pluginContextPtr *C.CPluginContext
if storagePluginContext != nil {
ckey := C.CString(storagePluginContext.EncryptionKey)
defer C.free(unsafe.Pointer(ckey))
var pluginContext C.CPluginContext
pluginContext.ez_id = C.int64_t(storagePluginContext.EncryptionZoneId)
pluginContext.collection_id = C.int64_t(storagePluginContext.CollectionId)
pluginContext.key = ckey
pluginContextPtr = &pluginContext
}
if storageConfig != nil {
cStorageConfig := C.CStorageConfig{
address: C.CString(storageConfig.GetAddress()),
bucket_name: C.CString(storageConfig.GetBucketName()),
access_key_id: C.CString(storageConfig.GetAccessKeyID()),
access_key_value: C.CString(storageConfig.GetSecretAccessKey()),
root_path: C.CString(storageConfig.GetRootPath()),
storage_type: C.CString(storageConfig.GetStorageType()),
cloud_provider: C.CString(storageConfig.GetCloudProvider()),
iam_endpoint: C.CString(storageConfig.GetIAMEndpoint()),
log_level: C.CString("Warn"), // TODO use config after storage support lower case configuration
useSSL: C.bool(storageConfig.GetUseSSL()),
sslCACert: C.CString(storageConfig.GetSslCACert()),
useIAM: C.bool(storageConfig.GetUseIAM()),
region: C.CString(storageConfig.GetRegion()),
useVirtualHost: C.bool(storageConfig.GetUseVirtualHost()),
requestTimeoutMs: C.int64_t(storageConfig.GetRequestTimeoutMs()),
gcp_credential_json: C.CString(storageConfig.GetGcpCredentialJSON()),
use_custom_part_upload: true,
max_connections: C.uint32_t(storageConfig.GetMaxConnections()),
}
defer C.free(unsafe.Pointer(cStorageConfig.address))
defer C.free(unsafe.Pointer(cStorageConfig.bucket_name))
defer C.free(unsafe.Pointer(cStorageConfig.access_key_id))
defer C.free(unsafe.Pointer(cStorageConfig.access_key_value))
defer C.free(unsafe.Pointer(cStorageConfig.root_path))
defer C.free(unsafe.Pointer(cStorageConfig.storage_type))
defer C.free(unsafe.Pointer(cStorageConfig.cloud_provider))
defer C.free(unsafe.Pointer(cStorageConfig.iam_endpoint))
defer C.free(unsafe.Pointer(cStorageConfig.log_level))
defer C.free(unsafe.Pointer(cStorageConfig.sslCACert))
defer C.free(unsafe.Pointer(cStorageConfig.region))
defer C.free(unsafe.Pointer(cStorageConfig.gcp_credential_json))
cNeededColumn := make([]*C.char, len(neededColumns))
for i, columnName := range neededColumns {
cNeededColumn[i] = C.CString(columnName)
defer C.free(unsafe.Pointer(cNeededColumn[i]))
}
cNeededColumnArray := (**C.char)(unsafe.Pointer(&cNeededColumn[0]))
cNumColumns := C.int64_t(len(neededColumns))
status = C.NewPackedFFIReaderWithManifest(cManifest, cSchema, cNeededColumnArray, cNumColumns, &cPackedReader, cStorageConfig, pluginContextPtr)
} else {
return nil, fmt.Errorf("storageConfig is required")
}
if err := ConsumeCStatusIntoError(&status); err != nil {
return nil, err
}
// Get the ArrowArrayStream
var cStream cdata.CArrowArrayStream
status = C.GetFFIReaderStream(cPackedReader, C.int64_t(8196), (*C.struct_ArrowArrayStream)(unsafe.Pointer(&cStream)))
if err := ConsumeCStatusIntoError(&status); err != nil {
C.CloseFFIReader(cPackedReader)
return nil, fmt.Errorf("failed to get reader stream: %w", err)
}
// Import the stream as a RecordReader
recordReader, err := cdata.ImportCRecordReader(&cStream, schema)
if err != nil {
C.CloseFFIReader(cPackedReader)
return nil, fmt.Errorf("failed to import record reader: %w", err)
}
return &FFIPackedReader{
cPackedReader: cPackedReader,
recordReader: recordReader,
schema: schema,
}, nil
}
// ReadNext reads the next record batch from the reader
func (r *FFIPackedReader) ReadNext() (arrow.Record, error) {
if r.recordReader == nil {
return nil, io.EOF
}
// no need to manual release
// stream reader will release previous one
// Read next record from the stream
rec, err := r.recordReader.Read()
if err != nil {
if err == io.EOF {
return nil, io.EOF
}
return nil, fmt.Errorf("failed to read next record: %w", err)
}
return rec, nil
}
// Close closes the FFI reader
func (r *FFIPackedReader) Close() error {
// no need to manual release current batch
// stream reader handles it
if r.recordReader != nil {
r.recordReader = nil
}
if r.cPackedReader != 0 {
status := C.CloseFFIReader(r.cPackedReader)
r.cPackedReader = 0
return ConsumeCStatusIntoError(&status)
}
return nil
}
// Schema returns the schema of the reader
func (r *FFIPackedReader) Schema() *arrow.Schema {
return r.schema
}
// Retain increases the reference count
func (r *FFIPackedReader) Retain() {
// if r.recordReader != nil {
// r.recordReader.Retain()
// }
}
// Release decreases the reference count
func (r *FFIPackedReader) Release() {
r.Close()
}
func GetManifest(manifestPath string, storageConfig *indexpb.StorageConfig) (manifest string, err error) {
basePath, version, err := UnmarshalManfestPath(manifestPath)
if err != nil {
return "", err
}
log.Info("GetManifest", zap.String("manifestPath", manifestPath), zap.String("basePath", basePath), zap.Int64("version", version))
cProperties, err := MakePropertiesFromStorageConfig(storageConfig, nil)
if err != nil {
return "", err
}
cBasePath := C.CString(basePath)
defer C.free(unsafe.Pointer(cBasePath))
var cManifest *C.char
var cVersion C.int64_t
result := C.get_latest_column_groups(cBasePath, cProperties, &cManifest, &cVersion)
err = HandleFFIResult(result)
if err != nil {
return "", err
}
manifest = C.GoString(cManifest)
return manifest, nil
}
// Ensure FFIPackedReader implements array.RecordReader interface
// var _ array.RecordReader = (*FFIPackedReader)(nil)