milvus/internal/datacoord/import_checker.go
yihao.dai 0fc2a4aa53
enhance: Optimize import scheduling and add time cost metric (#36601)
1. Optimize import scheduling strategic:
a. Revise slot weights, calculating them based on the number of files
and segments for both import and pre-import tasks.
b. Ensure that the DN executes tasks in ascending order of task ID.
2. Add time cost metric and log.

issue: https://github.com/milvus-io/milvus/issues/36600,
https://github.com/milvus-io/milvus/issues/36518

---------

Signed-off-by: bigsheeper <yihao.dai@zilliz.com>
2024-10-09 14:41:20 +08:00

524 lines
18 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 datacoord
import (
"context"
"fmt"
"sync"
"time"
"github.com/samber/lo"
"go.uber.org/zap"
"github.com/milvus-io/milvus/internal/datacoord/allocator"
"github.com/milvus-io/milvus/internal/datacoord/broker"
"github.com/milvus-io/milvus/internal/proto/datapb"
"github.com/milvus-io/milvus/internal/proto/indexpb"
"github.com/milvus-io/milvus/internal/proto/internalpb"
"github.com/milvus-io/milvus/internal/util/importutilv2"
"github.com/milvus-io/milvus/pkg/log"
"github.com/milvus-io/milvus/pkg/metrics"
"github.com/milvus-io/milvus/pkg/util/tsoutil"
)
type ImportChecker interface {
Start()
Close()
}
type importChecker struct {
meta *meta
broker broker.Broker
cluster Cluster
alloc allocator.Allocator
imeta ImportMeta
sjm StatsJobManager
closeOnce sync.Once
closeChan chan struct{}
}
func NewImportChecker(meta *meta,
broker broker.Broker,
cluster Cluster,
alloc allocator.Allocator,
imeta ImportMeta,
sjm StatsJobManager,
) ImportChecker {
return &importChecker{
meta: meta,
broker: broker,
cluster: cluster,
alloc: alloc,
imeta: imeta,
sjm: sjm,
closeChan: make(chan struct{}),
}
}
func (c *importChecker) Start() {
log.Info("start import checker")
var (
ticker1 = time.NewTicker(Params.DataCoordCfg.ImportCheckIntervalHigh.GetAsDuration(time.Second)) // 2s
ticker2 = time.NewTicker(Params.DataCoordCfg.ImportCheckIntervalLow.GetAsDuration(time.Second)) // 2min
)
defer ticker1.Stop()
defer ticker2.Stop()
for {
select {
case <-c.closeChan:
log.Info("import checker exited")
return
case <-ticker1.C:
jobs := c.imeta.GetJobBy()
for _, job := range jobs {
switch job.GetState() {
case internalpb.ImportJobState_Pending:
c.checkPendingJob(job)
case internalpb.ImportJobState_PreImporting:
c.checkPreImportingJob(job)
case internalpb.ImportJobState_Importing:
c.checkImportingJob(job)
case internalpb.ImportJobState_Stats:
c.checkStatsJob(job)
case internalpb.ImportJobState_IndexBuilding:
c.checkIndexBuildingJob(job)
case internalpb.ImportJobState_Failed:
c.checkFailedJob(job)
}
}
case <-ticker2.C:
jobs := c.imeta.GetJobBy()
for _, job := range jobs {
c.tryTimeoutJob(job)
c.checkGC(job)
}
jobsByColl := lo.GroupBy(jobs, func(job ImportJob) int64 {
return job.GetCollectionID()
})
for collID, collJobs := range jobsByColl {
c.checkCollection(collID, collJobs)
}
c.LogStats()
}
}
}
func (c *importChecker) Close() {
c.closeOnce.Do(func() {
close(c.closeChan)
})
}
func (c *importChecker) LogStats() {
logFunc := func(tasks []ImportTask, taskType TaskType) {
byState := lo.GroupBy(tasks, func(t ImportTask) datapb.ImportTaskStateV2 {
return t.GetState()
})
pending := len(byState[datapb.ImportTaskStateV2_Pending])
inProgress := len(byState[datapb.ImportTaskStateV2_InProgress])
completed := len(byState[datapb.ImportTaskStateV2_Completed])
failed := len(byState[datapb.ImportTaskStateV2_Failed])
log.Info("import task stats", zap.String("type", taskType.String()),
zap.Int("pending", pending), zap.Int("inProgress", inProgress),
zap.Int("completed", completed), zap.Int("failed", failed))
metrics.ImportTasks.WithLabelValues(taskType.String(), datapb.ImportTaskStateV2_Pending.String()).Set(float64(pending))
metrics.ImportTasks.WithLabelValues(taskType.String(), datapb.ImportTaskStateV2_InProgress.String()).Set(float64(inProgress))
metrics.ImportTasks.WithLabelValues(taskType.String(), datapb.ImportTaskStateV2_Completed.String()).Set(float64(completed))
metrics.ImportTasks.WithLabelValues(taskType.String(), datapb.ImportTaskStateV2_Failed.String()).Set(float64(failed))
}
tasks := c.imeta.GetTaskBy(WithType(PreImportTaskType))
logFunc(tasks, PreImportTaskType)
tasks = c.imeta.GetTaskBy(WithType(ImportTaskType))
logFunc(tasks, ImportTaskType)
}
func (c *importChecker) getLackFilesForPreImports(job ImportJob) []*internalpb.ImportFile {
lacks := lo.KeyBy(job.GetFiles(), func(file *internalpb.ImportFile) int64 {
return file.GetId()
})
exists := c.imeta.GetTaskBy(WithType(PreImportTaskType), WithJob(job.GetJobID()))
for _, task := range exists {
for _, file := range task.GetFileStats() {
delete(lacks, file.GetImportFile().GetId())
}
}
return lo.Values(lacks)
}
func (c *importChecker) getLackFilesForImports(job ImportJob) []*datapb.ImportFileStats {
preimports := c.imeta.GetTaskBy(WithType(PreImportTaskType), WithJob(job.GetJobID()))
lacks := make(map[int64]*datapb.ImportFileStats, 0)
for _, t := range preimports {
if t.GetState() != datapb.ImportTaskStateV2_Completed {
// Preimport tasks are not fully completed, thus generating imports should not be triggered.
return nil
}
for _, stat := range t.GetFileStats() {
lacks[stat.GetImportFile().GetId()] = stat
}
}
exists := c.imeta.GetTaskBy(WithType(ImportTaskType), WithJob(job.GetJobID()))
for _, task := range exists {
for _, file := range task.GetFileStats() {
delete(lacks, file.GetImportFile().GetId())
}
}
return lo.Values(lacks)
}
func (c *importChecker) checkPendingJob(job ImportJob) {
log := log.With(zap.Int64("jobID", job.GetJobID()))
lacks := c.getLackFilesForPreImports(job)
if len(lacks) == 0 {
return
}
fileGroups := lo.Chunk(lacks, Params.DataCoordCfg.FilesPerPreImportTask.GetAsInt())
newTasks, err := NewPreImportTasks(fileGroups, job, c.alloc)
if err != nil {
log.Warn("new preimport tasks failed", zap.Error(err))
return
}
for _, t := range newTasks {
err = c.imeta.AddTask(t)
if err != nil {
log.Warn("add preimport task failed", WrapTaskLog(t, zap.Error(err))...)
return
}
log.Info("add new preimport task", WrapTaskLog(t)...)
}
err = c.imeta.UpdateJob(job.GetJobID(), UpdateJobState(internalpb.ImportJobState_PreImporting))
if err != nil {
log.Warn("failed to update job state to PreImporting", zap.Error(err))
return
}
pendingDuration := job.GetTR().RecordSpan()
metrics.ImportJobLatency.WithLabelValues(metrics.ImportStagePending).Observe(float64(pendingDuration.Milliseconds()))
log.Info("import job start to execute", zap.Duration("jobTimeCost/pending", pendingDuration))
}
func (c *importChecker) checkPreImportingJob(job ImportJob) {
log := log.With(zap.Int64("jobID", job.GetJobID()))
lacks := c.getLackFilesForImports(job)
if len(lacks) == 0 {
return
}
requestSize, err := CheckDiskQuota(job, c.meta, c.imeta)
if err != nil {
log.Warn("import failed, disk quota exceeded", zap.Error(err))
err = c.imeta.UpdateJob(job.GetJobID(), UpdateJobState(internalpb.ImportJobState_Failed), UpdateJobReason(err.Error()))
if err != nil {
log.Warn("failed to update job state to Failed", zap.Error(err))
}
return
}
allDiskIndex := c.meta.indexMeta.AreAllDiskIndex(job.GetCollectionID(), job.GetSchema())
groups := RegroupImportFiles(job, lacks, allDiskIndex)
newTasks, err := NewImportTasks(groups, job, c.alloc, c.meta)
if err != nil {
log.Warn("new import tasks failed", zap.Error(err))
return
}
for _, t := range newTasks {
err = c.imeta.AddTask(t)
if err != nil {
log.Warn("add new import task failed", WrapTaskLog(t, zap.Error(err))...)
return
}
log.Info("add new import task", WrapTaskLog(t)...)
}
err = c.imeta.UpdateJob(job.GetJobID(), UpdateJobState(internalpb.ImportJobState_Importing), UpdateRequestedDiskSize(requestSize))
if err != nil {
log.Warn("failed to update job state to Importing", zap.Error(err))
return
}
preImportDuration := job.GetTR().RecordSpan()
metrics.ImportJobLatency.WithLabelValues(metrics.ImportStagePreImport).Observe(float64(preImportDuration.Milliseconds()))
log.Info("import job preimport done", zap.Duration("jobTimeCost/preimport", preImportDuration))
}
func (c *importChecker) checkImportingJob(job ImportJob) {
log := log.With(zap.Int64("jobID", job.GetJobID()))
tasks := c.imeta.GetTaskBy(WithType(ImportTaskType), WithJob(job.GetJobID()))
for _, t := range tasks {
if t.GetState() != datapb.ImportTaskStateV2_Completed {
return
}
}
err := c.imeta.UpdateJob(job.GetJobID(), UpdateJobState(internalpb.ImportJobState_Stats))
if err != nil {
log.Warn("failed to update job state to Stats", zap.Error(err))
return
}
importDuration := job.GetTR().RecordSpan()
metrics.ImportJobLatency.WithLabelValues(metrics.ImportStageImport).Observe(float64(importDuration.Milliseconds()))
log.Info("import job import done", zap.Duration("jobTimeCost/import", importDuration))
}
func (c *importChecker) checkStatsJob(job ImportJob) {
log := log.With(zap.Int64("jobID", job.GetJobID()))
updateJobState := func(state internalpb.ImportJobState) {
err := c.imeta.UpdateJob(job.GetJobID(), UpdateJobState(state))
if err != nil {
log.Warn("failed to update job state", zap.Error(err))
return
}
statsDuration := job.GetTR().RecordSpan()
metrics.ImportJobLatency.WithLabelValues(metrics.ImportStageStats).Observe(float64(statsDuration.Milliseconds()))
log.Info("import job stats done", zap.Duration("jobTimeCost/stats", statsDuration))
}
// Skip stats stage if not enable stats or is l0 import.
if !Params.DataCoordCfg.EnableStatsTask.GetAsBool() || importutilv2.IsL0Import(job.GetOptions()) {
updateJobState(internalpb.ImportJobState_IndexBuilding)
return
}
// Check and trigger stats tasks.
var (
taskCnt = 0
doneCnt = 0
)
tasks := c.imeta.GetTaskBy(WithType(ImportTaskType), WithJob(job.GetJobID()))
for _, task := range tasks {
originSegmentIDs := task.(*importTask).GetSegmentIDs()
statsSegmentIDs := task.(*importTask).GetStatsSegmentIDs()
taskCnt += len(originSegmentIDs)
for i, originSegmentID := range originSegmentIDs {
taskLogFields := WrapTaskLog(task, zap.Int64("origin", originSegmentID), zap.Int64("stats", statsSegmentIDs[i]))
state := c.sjm.GetStatsTaskState(originSegmentID, indexpb.StatsSubJob_Sort)
switch state {
case indexpb.JobState_JobStateNone:
err := c.sjm.SubmitStatsTask(originSegmentID, statsSegmentIDs[i], indexpb.StatsSubJob_Sort, false)
if err != nil {
log.Warn("submit stats task failed", zap.Error(err))
continue
}
log.Info("submit stats task done", taskLogFields...)
case indexpb.JobState_JobStateInit, indexpb.JobState_JobStateRetry, indexpb.JobState_JobStateInProgress:
log.Debug("waiting for stats task...", taskLogFields...)
case indexpb.JobState_JobStateFailed:
log.Warn("import job stats failed", taskLogFields...)
updateJobState(internalpb.ImportJobState_Failed)
return
case indexpb.JobState_JobStateFinished:
doneCnt++
}
}
}
// All segments are stats-ed. Update job state to `IndexBuilding`.
if taskCnt == doneCnt {
updateJobState(internalpb.ImportJobState_IndexBuilding)
}
}
func (c *importChecker) checkIndexBuildingJob(job ImportJob) {
log := log.With(zap.Int64("jobID", job.GetJobID()))
tasks := c.imeta.GetTaskBy(WithType(ImportTaskType), WithJob(job.GetJobID()))
originSegmentIDs := lo.FlatMap(tasks, func(t ImportTask, _ int) []int64 {
return t.(*importTask).GetSegmentIDs()
})
statsSegmentIDs := lo.FlatMap(tasks, func(t ImportTask, _ int) []int64 {
return t.(*importTask).GetStatsSegmentIDs()
})
targetSegmentIDs := statsSegmentIDs
if !Params.DataCoordCfg.EnableStatsTask.GetAsBool() {
targetSegmentIDs = originSegmentIDs
}
unindexed := c.meta.indexMeta.GetUnindexedSegments(job.GetCollectionID(), targetSegmentIDs)
if Params.DataCoordCfg.WaitForIndex.GetAsBool() && len(unindexed) > 0 && !importutilv2.IsL0Import(job.GetOptions()) {
for _, segmentID := range unindexed {
select {
case getBuildIndexChSingleton() <- segmentID: // accelerate index building:
default:
}
}
log.Debug("waiting for import segments building index...", zap.Int64s("unindexed", unindexed))
return
}
buildIndexDuration := job.GetTR().RecordSpan()
metrics.ImportJobLatency.WithLabelValues(metrics.ImportStageBuildIndex).Observe(float64(buildIndexDuration.Milliseconds()))
log.Info("import job build index done", zap.Duration("jobTimeCost/buildIndex", buildIndexDuration))
// Here, all segment indexes have been successfully built, try unset isImporting flag for all segments.
isImportingSegments := lo.Filter(append(originSegmentIDs, statsSegmentIDs...), func(segmentID int64, _ int) bool {
segment := c.meta.GetSegment(segmentID)
if segment == nil {
log.Warn("cannot find segment", zap.Int64("segmentID", segmentID))
return false
}
return segment.GetIsImporting()
})
channels, err := c.meta.GetSegmentsChannels(isImportingSegments)
if err != nil {
log.Warn("get segments channels failed", zap.Error(err))
return
}
for _, segmentID := range isImportingSegments {
channelCP := c.meta.GetChannelCheckpoint(channels[segmentID])
if channelCP == nil {
log.Warn("nil channel checkpoint")
return
}
op1 := UpdateStartPosition([]*datapb.SegmentStartPosition{{StartPosition: channelCP, SegmentID: segmentID}})
op2 := UpdateDmlPosition(segmentID, channelCP)
op3 := UpdateIsImporting(segmentID, false)
err = c.meta.UpdateSegmentsInfo(op1, op2, op3)
if err != nil {
log.Warn("update import segment failed", zap.Error(err))
return
}
}
// all finished, update import job state to `Completed`.
completeTime := time.Now().Format("2006-01-02T15:04:05Z07:00")
err = c.imeta.UpdateJob(job.GetJobID(), UpdateJobState(internalpb.ImportJobState_Completed), UpdateJobCompleteTime(completeTime))
if err != nil {
log.Warn("failed to update job state to Completed", zap.Error(err))
return
}
totalDuration := job.GetTR().ElapseSpan()
metrics.ImportJobLatency.WithLabelValues(metrics.TotalLabel).Observe(float64(totalDuration.Milliseconds()))
log.Info("import job all completed", zap.Duration("jobTimeCost/total", totalDuration))
}
func (c *importChecker) checkFailedJob(job ImportJob) {
tasks := c.imeta.GetTaskBy(WithType(ImportTaskType), WithJob(job.GetJobID()))
originSegmentIDs := lo.FlatMap(tasks, func(t ImportTask, _ int) []int64 {
return t.(*importTask).GetSegmentIDs()
})
for _, originSegmentID := range originSegmentIDs {
err := c.sjm.DropStatsTask(originSegmentID, indexpb.StatsSubJob_Sort)
if err != nil {
log.Warn("Drop stats task failed", zap.Int64("jobID", job.GetJobID()))
return
}
}
c.tryFailingTasks(job)
}
func (c *importChecker) tryFailingTasks(job ImportJob) {
tasks := c.imeta.GetTaskBy(WithJob(job.GetJobID()), WithStates(datapb.ImportTaskStateV2_Pending,
datapb.ImportTaskStateV2_InProgress, datapb.ImportTaskStateV2_Completed))
if len(tasks) == 0 {
return
}
log.Warn("Import job has failed, all tasks with the same jobID will be marked as failed",
zap.Int64("jobID", job.GetJobID()), zap.String("reason", job.GetReason()))
for _, task := range tasks {
err := c.imeta.UpdateTask(task.GetTaskID(), UpdateState(datapb.ImportTaskStateV2_Failed),
UpdateReason(job.GetReason()))
if err != nil {
log.Warn("failed to update import task state to failed", WrapTaskLog(task, zap.Error(err))...)
continue
}
}
}
func (c *importChecker) tryTimeoutJob(job ImportJob) {
timeoutTime := tsoutil.PhysicalTime(job.GetTimeoutTs())
if time.Now().After(timeoutTime) {
log.Warn("Import timeout, expired the specified time limit",
zap.Int64("jobID", job.GetJobID()), zap.Time("timeoutTime", timeoutTime))
err := c.imeta.UpdateJob(job.GetJobID(), UpdateJobState(internalpb.ImportJobState_Failed),
UpdateJobReason("import timeout"))
if err != nil {
log.Warn("failed to update job state to Failed", zap.Int64("jobID", job.GetJobID()), zap.Error(err))
}
}
}
func (c *importChecker) checkCollection(collectionID int64, jobs []ImportJob) {
if len(jobs) == 0 {
return
}
ctx, cancel := context.WithTimeout(context.Background(), 10*time.Second)
defer cancel()
has, err := c.broker.HasCollection(ctx, collectionID)
if err != nil {
log.Warn("verify existence of collection failed", zap.Int64("collection", collectionID), zap.Error(err))
return
}
if !has {
jobs = lo.Filter(jobs, func(job ImportJob, _ int) bool {
return job.GetState() != internalpb.ImportJobState_Failed
})
for _, job := range jobs {
err = c.imeta.UpdateJob(job.GetJobID(), UpdateJobState(internalpb.ImportJobState_Failed),
UpdateJobReason(fmt.Sprintf("collection %d dropped", collectionID)))
if err != nil {
log.Warn("failed to update job state to Failed", zap.Int64("jobID", job.GetJobID()), zap.Error(err))
}
}
}
}
func (c *importChecker) checkGC(job ImportJob) {
if job.GetState() != internalpb.ImportJobState_Completed &&
job.GetState() != internalpb.ImportJobState_Failed {
return
}
cleanupTime := tsoutil.PhysicalTime(job.GetCleanupTs())
if time.Now().After(cleanupTime) {
log := log.With(zap.Int64("jobID", job.GetJobID()))
GCRetention := Params.DataCoordCfg.ImportTaskRetention.GetAsDuration(time.Second)
log.Info("job has reached the GC retention",
zap.Time("cleanupTime", cleanupTime), zap.Duration("GCRetention", GCRetention))
tasks := c.imeta.GetTaskBy(WithJob(job.GetJobID()))
shouldRemoveJob := true
for _, task := range tasks {
if job.GetState() == internalpb.ImportJobState_Failed && task.GetType() == ImportTaskType {
if len(task.(*importTask).GetSegmentIDs()) != 0 || len(task.(*importTask).GetStatsSegmentIDs()) != 0 {
shouldRemoveJob = false
continue
}
}
if task.GetNodeID() != NullNodeID {
shouldRemoveJob = false
continue
}
err := c.imeta.RemoveTask(task.GetTaskID())
if err != nil {
log.Warn("remove task failed during GC", WrapTaskLog(task, zap.Error(err))...)
shouldRemoveJob = false
continue
}
log.Info("reached GC retention, task removed", WrapTaskLog(task)...)
}
if !shouldRemoveJob {
return
}
err := c.imeta.RemoveJob(job.GetJobID())
if err != nil {
log.Warn("remove import job failed", zap.Error(err))
return
}
log.Info("import job removed")
}
}