yihao.dai 5e525eb3bf
enhance: Retry reads from object storage on rate limit error (#46455)
This PR improves the robustness of object storage operations by retrying
both explicit throttling errors (e.g. HTTP 429, SlowDown, ServerBusy).
These errors commonly occur under high concurrency and are typically
recoverable with bounded retries.

issue: https://github.com/milvus-io/milvus/issues/44772

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

* **New Features**
* Configurable retry support for reads from object storage and improved
mapping of transient/rate-limit errors.
* Added a retryable reader wrapper used by CSV/JSON/Parquet/Numpy import
paths.

* **Configuration**
  * New parameter to control storage read retry attempts.

* **Tests**
* Expanded unit tests covering error mapping and retry behaviors across
storage backends.
* Standardized mock readers and test initialization to simplify test
setups.

<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: bigsheeper <yihao.dai@zilliz.com>
2025-12-23 11:03:18 +08:00

271 lines
7.0 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 json
import (
"context"
"encoding/json"
"fmt"
"io"
"strings"
"go.uber.org/atomic"
"github.com/milvus-io/milvus-proto/go-api/v2/schemapb"
"github.com/milvus-io/milvus/internal/storage"
"github.com/milvus-io/milvus/internal/util/importutilv2/common"
"github.com/milvus-io/milvus/pkg/v2/util/merr"
)
const (
RowRootNode = "rows"
)
type Row = map[storage.FieldID]any
type reader struct {
ctx context.Context
cm storage.ChunkManager
cmr storage.FileReader
schema *schemapb.CollectionSchema
fileSize *atomic.Int64
filePath string
dec *json.Decoder
bufferSize int
count int64
isOldFormat bool
isLinesFormat bool
parser RowParser
}
func newReader(ctx context.Context, cm storage.ChunkManager, schema *schemapb.CollectionSchema, path string, bufferSize int, isLinesFormat bool) (*reader, error) {
r, err := cm.Reader(ctx, path)
if err != nil {
return nil, merr.WrapErrImportFailed(fmt.Sprintf("read json file failed, path=%s, err=%s", path, err.Error()))
}
retryableReader := common.NewRetryableReader(ctx, path, r)
count, err := common.EstimateReadCountPerBatch(bufferSize, schema)
if err != nil {
return nil, err
}
reader := &reader{
ctx: ctx,
cm: cm,
cmr: retryableReader,
schema: schema,
fileSize: atomic.NewInt64(0),
filePath: path,
dec: json.NewDecoder(retryableReader),
bufferSize: bufferSize,
count: count,
isLinesFormat: isLinesFormat,
}
reader.parser, err = NewRowParser(schema)
if err != nil {
return nil, err
}
err = reader.Init()
if err != nil {
return nil, err
}
return reader, nil
}
func NewReader(ctx context.Context, cm storage.ChunkManager, schema *schemapb.CollectionSchema, path string, bufferSize int) (*reader, error) {
reader, err := newReader(ctx, cm, schema, path, bufferSize, false)
return reader, err
}
func NewLinesReader(ctx context.Context, cm storage.ChunkManager, schema *schemapb.CollectionSchema, path string, bufferSize int) (*reader, error) {
reader, err := newReader(ctx, cm, schema, path, bufferSize, true)
return reader, err
}
func (j *reader) Init() error {
// Treat number value as a string instead of a float64.
// By default, json lib treat all number values as float64,
// but if an int64 value has more than 15 digits,
// the value would be incorrect after converting from float64.
j.dec.UseNumber()
// JSONLines format, each line is a JSON dict like this:
// a.jsonl
// {"a": 1}
// {"b": 2}
if j.isLinesFormat {
return nil
}
// Normal JSON format, accept two formats:
// old_format.json
// {
// "rows": [
// {"a": 1},
// {"b": 2}
// ]
// }
//
// list_format.json
// [
// {"a": 1},
// {"b": 2}
// ]
t, err := j.dec.Token()
if err != nil {
return merr.WrapErrImportFailed(fmt.Sprintf("init failed, failed to decode JSON, error: %v", err))
}
if t != json.Delim('{') && t != json.Delim('[') {
return merr.WrapErrImportFailed("invalid JSON format, the content should be started with '{' or '['")
}
j.isOldFormat = t == json.Delim('{')
return nil
}
func (j *reader) Read() (*storage.InsertData, error) {
insertData, err := storage.NewInsertData(j.schema)
if err != nil {
return nil, err
}
if !j.dec.More() {
return nil, io.EOF
}
if j.isLinesFormat {
if err = j.readJSONLines(insertData); err != nil {
return nil, err
}
} else {
if err = j.readNormalJSON(insertData); err != nil {
return nil, err
}
}
return insertData, nil
}
func (j *reader) readNormalJSON(insertData *storage.InsertData) error {
// In old versions(2.2/2.3), milvus requires users to create the JSON begins with a root key "rows"
// old_format.json
// {
// "rows": [
// {"a": 1},
// {"b": 2}
// ]
// }
// this section is to detect the old format
if j.isOldFormat {
// read the key
t, err := j.dec.Token()
if err != nil {
return merr.WrapErrImportFailed(fmt.Sprintf("failed to decode the JSON file, error: %v", err))
}
key := t.(string)
keyLower := strings.ToLower(key)
// the root key should be RowRootNode
if keyLower != RowRootNode {
return merr.WrapErrImportFailed(fmt.Sprintf("invalid JSON format, the root key should be '%s', but get '%s'", RowRootNode, key))
}
// started by '['
t, err = j.dec.Token()
if err != nil {
return merr.WrapErrImportFailed(fmt.Sprintf("failed to decode the JSON file, error: %v", err))
}
if t != json.Delim('[') {
return merr.WrapErrImportFailed("invalid JSON format, rows list should begin with '['")
}
j.isOldFormat = false
}
// Read the content under the key "rows"(or inside the list) line by line
if err := j.readJSONLines(insertData); err != nil {
return err
}
// Verify the content is ended with "]"
if !j.dec.More() {
t, err := j.dec.Token()
if err != nil {
return merr.WrapErrImportFailed(fmt.Sprintf("failed to decode JSON, error: %v", err))
}
if t != json.Delim(']') {
return merr.WrapErrImportFailed("invalid JSON format, rows list should end with ']'")
}
}
return nil
}
func (j *reader) readJSONLines(insertData *storage.InsertData) error {
// Read line by line, the json.Decoder.Decode() can ignore the comma at the end of each line,
// both of the two formats will get the same result:
// 1. lines ending with a comma:
// {"a": 1},
// {"b": 2},
// {"c": 3}
// 2. lines ending without comma:
// {"a": 1}
// {"b": 2}
// {"c": 3}
var cnt int64 = 0
for j.dec.More() {
var value any
if err := j.dec.Decode(&value); err != nil {
return merr.WrapErrImportFailed(fmt.Sprintf("failed to decode row, error: %v", err))
}
row, err := j.parser.Parse(value)
if err != nil {
return merr.WrapErrImportFailed(fmt.Sprintf("failed to parse row, error: %v", err))
}
err = insertData.Append(row)
if err != nil {
return merr.WrapErrImportFailed(fmt.Sprintf("failed to append row, err=%s", err.Error()))
}
cnt++
if cnt >= j.count {
cnt = 0
if insertData.GetMemorySize() >= j.bufferSize {
break
}
}
}
return nil
}
func (j *reader) Size() (int64, error) {
if size := j.fileSize.Load(); size != 0 {
return size, nil
}
size, err := j.cm.Size(j.ctx, j.filePath)
if err != nil {
return 0, err
}
j.fileSize.Store(size)
return size, nil
}
func (j *reader) Close() {
if j.cmr != nil {
j.cmr.Close()
}
}