githaven/modules/indexer/code/indexer.go
Jason Song 375fd15fbf
Refactor indexer (#25174)
Refactor `modules/indexer` to make it more maintainable. And it can be
easier to support more features. I'm trying to solve some of issue
searching, this is a precursor to making functional changes.

Current supported engines and the index versions:

| engines | issues | code |
| - | - | - |
| db | Just a wrapper for database queries, doesn't need version | - |
| bleve | The version of index is **2** | The version of index is **6**
|
| elasticsearch | The old index has no version, will be treated as
version **0** in this PR | The version of index is **1** |
| meilisearch | The old index has no version, will be treated as version
**0** in this PR | - |


## Changes

### Split

Splited it into mutiple packages

```text
indexer
├── internal
│   ├── bleve
│   ├── db
│   ├── elasticsearch
│   └── meilisearch
├── code
│   ├── bleve
│   ├── elasticsearch
│   └── internal
└── issues
    ├── bleve
    ├── db
    ├── elasticsearch
    ├── internal
    └── meilisearch
```

- `indexer/interanal`: Internal shared package for indexer.
- `indexer/interanal/[engine]`: Internal shared package for each engine
(bleve/db/elasticsearch/meilisearch).
- `indexer/code`: Implementations for code indexer.
- `indexer/code/internal`: Internal shared package for code indexer.
- `indexer/code/[engine]`: Implementation via each engine for code
indexer.
- `indexer/issues`: Implementations for issues indexer.

### Deduplication

- Combine `Init/Ping/Close` for code indexer and issues indexer.
- ~Combine `issues.indexerHolder` and `code.wrappedIndexer` to
`internal.IndexHolder`.~ Remove it, use dummy indexer instead when the
indexer is not ready.
- Duplicate two copies of creating ES clients.
- Duplicate two copies of `indexerID()`.


### Enhancement

- [x] Support index version for elasticsearch issues indexer, the old
index without version will be treated as version 0.
- [x] Fix spell of `elastic_search/ElasticSearch`, it should be
`Elasticsearch`.
- [x] Improve versioning of ES index. We don't need `Aliases`:
- Gitea does't need aliases for "Zero Downtime" because it never delete
old indexes.
- The old code of issues indexer uses the orignal name to create issue
index, so it's tricky to convert it to an alias.
- [x] Support index version for meilisearch issues indexer, the old
index without version will be treated as version 0.
- [x] Do "ping" only when `Ping` has been called, don't ping
periodically and cache the status.
- [x] Support the context parameter whenever possible.
- [x] Fix outdated example config.
- [x] Give up the requeue logic of issues indexer: When indexing fails,
call Ping to check if it was caused by the engine being unavailable, and
only requeue the task if the engine is unavailable.
- It is fragile and tricky, could cause data losing (It did happen when
I was doing some tests for this PR). And it works for ES only.
- Just always requeue the failed task, if it caused by bad data, it's a
bug of Gitea which should be fixed.

---------

Co-authored-by: Giteabot <teabot@gitea.io>
2023-06-23 12:37:56 +00:00

329 lines
10 KiB
Go

// Copyright 2016 The Gitea Authors. All rights reserved.
// SPDX-License-Identifier: MIT
package code
import (
"context"
"os"
"runtime/pprof"
"sync/atomic"
"time"
"code.gitea.io/gitea/models/db"
repo_model "code.gitea.io/gitea/models/repo"
"code.gitea.io/gitea/modules/graceful"
"code.gitea.io/gitea/modules/indexer/code/bleve"
"code.gitea.io/gitea/modules/indexer/code/elasticsearch"
"code.gitea.io/gitea/modules/indexer/code/internal"
"code.gitea.io/gitea/modules/log"
"code.gitea.io/gitea/modules/process"
"code.gitea.io/gitea/modules/queue"
"code.gitea.io/gitea/modules/setting"
"code.gitea.io/gitea/modules/util"
)
var (
indexerQueue *queue.WorkerPoolQueue[*internal.IndexerData]
// globalIndexer is the global indexer, it cannot be nil.
// When the real indexer is not ready, it will be a dummy indexer which will return error to explain it's not ready.
// So it's always safe use it as *globalIndexer.Load() and call its methods.
globalIndexer atomic.Pointer[internal.Indexer]
dummyIndexer *internal.Indexer
)
func init() {
i := internal.NewDummyIndexer()
dummyIndexer = &i
globalIndexer.Store(dummyIndexer)
}
func index(ctx context.Context, indexer internal.Indexer, repoID int64) error {
repo, err := repo_model.GetRepositoryByID(ctx, repoID)
if repo_model.IsErrRepoNotExist(err) {
return indexer.Delete(ctx, repoID)
}
if err != nil {
return err
}
repoTypes := setting.Indexer.RepoIndexerRepoTypes
if len(repoTypes) == 0 {
repoTypes = []string{"sources"}
}
// skip forks from being indexed if unit is not present
if !util.SliceContains(repoTypes, "forks") && repo.IsFork {
return nil
}
// skip mirrors from being indexed if unit is not present
if !util.SliceContains(repoTypes, "mirrors") && repo.IsMirror {
return nil
}
// skip templates from being indexed if unit is not present
if !util.SliceContains(repoTypes, "templates") && repo.IsTemplate {
return nil
}
// skip regular repos from being indexed if unit is not present
if !util.SliceContains(repoTypes, "sources") && !repo.IsFork && !repo.IsMirror && !repo.IsTemplate {
return nil
}
sha, err := getDefaultBranchSha(ctx, repo)
if err != nil {
return err
}
changes, err := getRepoChanges(ctx, repo, sha)
if err != nil {
return err
} else if changes == nil {
return nil
}
if err := indexer.Index(ctx, repo, sha, changes); err != nil {
return err
}
return repo_model.UpdateIndexerStatus(ctx, repo, repo_model.RepoIndexerTypeCode, sha)
}
// Init initialize the repo indexer
func Init() {
if !setting.Indexer.RepoIndexerEnabled {
(*globalIndexer.Load()).Close()
return
}
ctx, cancel, finished := process.GetManager().AddTypedContext(context.Background(), "Service: CodeIndexer", process.SystemProcessType, false)
graceful.GetManager().RunAtTerminate(func() {
select {
case <-ctx.Done():
return
default:
}
cancel()
log.Debug("Closing repository indexer")
(*globalIndexer.Load()).Close()
log.Info("PID: %d Repository Indexer closed", os.Getpid())
finished()
})
waitChannel := make(chan time.Duration, 1)
// Create the Queue
switch setting.Indexer.RepoType {
case "bleve", "elasticsearch":
handler := func(items ...*internal.IndexerData) (unhandled []*internal.IndexerData) {
indexer := *globalIndexer.Load()
for _, indexerData := range items {
log.Trace("IndexerData Process Repo: %d", indexerData.RepoID)
// FIXME: it seems there is a bug in `CatFileBatch` or `nio.Pipe`, which will cause the process to hang forever in rare cases
/*
sync.(*Cond).Wait(cond.go:70)
github.com/djherbis/nio/v3.(*PipeReader).Read(sync.go:106)
bufio.(*Reader).fill(bufio.go:106)
bufio.(*Reader).ReadSlice(bufio.go:372)
bufio.(*Reader).collectFragments(bufio.go:447)
bufio.(*Reader).ReadString(bufio.go:494)
code.gitea.io/gitea/modules/git.ReadBatchLine(batch_reader.go:149)
code.gitea.io/gitea/modules/indexer/code.(*BleveIndexer).addUpdate(bleve.go:214)
code.gitea.io/gitea/modules/indexer/code.(*BleveIndexer).Index(bleve.go:296)
code.gitea.io/gitea/modules/indexer/code.(*wrappedIndexer).Index(wrapped.go:74)
code.gitea.io/gitea/modules/indexer/code.index(indexer.go:105)
*/
if err := index(ctx, indexer, indexerData.RepoID); err != nil {
unhandled = append(unhandled, indexerData)
if !setting.IsInTesting {
log.Error("Codes indexer handler: index error for repo %v: %v", indexerData.RepoID, err)
}
}
}
return unhandled
}
indexerQueue = queue.CreateUniqueQueue(ctx, "code_indexer", handler)
if indexerQueue == nil {
log.Fatal("Unable to create codes indexer queue")
}
default:
log.Fatal("Unknown codes indexer type; %s", setting.Indexer.RepoType)
}
go func() {
pprof.SetGoroutineLabels(ctx)
start := time.Now()
var (
rIndexer internal.Indexer
existed bool
err error
)
switch setting.Indexer.RepoType {
case "bleve":
log.Info("PID: %d Initializing Repository Indexer at: %s", os.Getpid(), setting.Indexer.RepoPath)
defer func() {
if err := recover(); err != nil {
log.Error("PANIC whilst initializing repository indexer: %v\nStacktrace: %s", err, log.Stack(2))
log.Error("The indexer files are likely corrupted and may need to be deleted")
log.Error("You can completely remove the \"%s\" directory to make Gitea recreate the indexes", setting.Indexer.RepoPath)
}
}()
rIndexer = bleve.NewIndexer(setting.Indexer.RepoPath)
existed, err = rIndexer.Init(ctx)
if err != nil {
cancel()
(*globalIndexer.Load()).Close()
close(waitChannel)
log.Fatal("PID: %d Unable to initialize the bleve Repository Indexer at path: %s Error: %v", os.Getpid(), setting.Indexer.RepoPath, err)
}
case "elasticsearch":
log.Info("PID: %d Initializing Repository Indexer at: %s", os.Getpid(), setting.Indexer.RepoConnStr)
defer func() {
if err := recover(); err != nil {
log.Error("PANIC whilst initializing repository indexer: %v\nStacktrace: %s", err, log.Stack(2))
log.Error("The indexer files are likely corrupted and may need to be deleted")
log.Error("You can completely remove the \"%s\" index to make Gitea recreate the indexes", setting.Indexer.RepoConnStr)
}
}()
rIndexer = elasticsearch.NewIndexer(setting.Indexer.RepoConnStr, setting.Indexer.RepoIndexerName)
if err != nil {
cancel()
(*globalIndexer.Load()).Close()
close(waitChannel)
log.Fatal("PID: %d Unable to create the elasticsearch Repository Indexer connstr: %s Error: %v", os.Getpid(), setting.Indexer.RepoConnStr, err)
}
existed, err = rIndexer.Init(ctx)
if err != nil {
cancel()
(*globalIndexer.Load()).Close()
close(waitChannel)
log.Fatal("PID: %d Unable to initialize the elasticsearch Repository Indexer connstr: %s Error: %v", os.Getpid(), setting.Indexer.RepoConnStr, err)
}
default:
log.Fatal("PID: %d Unknown Indexer type: %s", os.Getpid(), setting.Indexer.RepoType)
}
globalIndexer.Store(&rIndexer)
// Start processing the queue
go graceful.GetManager().RunWithCancel(indexerQueue)
if !existed { // populate the index because it's created for the first time
go graceful.GetManager().RunWithShutdownContext(populateRepoIndexer)
}
select {
case waitChannel <- time.Since(start):
case <-graceful.GetManager().IsShutdown():
}
close(waitChannel)
}()
if setting.Indexer.StartupTimeout > 0 {
go func() {
pprof.SetGoroutineLabels(ctx)
timeout := setting.Indexer.StartupTimeout
if graceful.GetManager().IsChild() && setting.GracefulHammerTime > 0 {
timeout += setting.GracefulHammerTime
}
select {
case <-graceful.GetManager().IsShutdown():
log.Warn("Shutdown before Repository Indexer completed initialization")
cancel()
(*globalIndexer.Load()).Close()
case duration, ok := <-waitChannel:
if !ok {
log.Warn("Repository Indexer Initialization failed")
cancel()
(*globalIndexer.Load()).Close()
return
}
log.Info("Repository Indexer Initialization took %v", duration)
case <-time.After(timeout):
cancel()
(*globalIndexer.Load()).Close()
log.Fatal("Repository Indexer Initialization Timed-Out after: %v", timeout)
}
}()
}
}
// UpdateRepoIndexer update a repository's entries in the indexer
func UpdateRepoIndexer(repo *repo_model.Repository) {
indexData := &internal.IndexerData{RepoID: repo.ID}
if err := indexerQueue.Push(indexData); err != nil {
log.Error("Update repo index data %v failed: %v", indexData, err)
}
}
// IsAvailable checks if issue indexer is available
func IsAvailable(ctx context.Context) bool {
return (*globalIndexer.Load()).Ping(ctx) == nil
}
// populateRepoIndexer populate the repo indexer with pre-existing data. This
// should only be run when the indexer is created for the first time.
func populateRepoIndexer(ctx context.Context) {
log.Info("Populating the repo indexer with existing repositories")
exist, err := db.IsTableNotEmpty("repository")
if err != nil {
log.Fatal("System error: %v", err)
} else if !exist {
return
}
// if there is any existing repo indexer metadata in the DB, delete it
// since we are starting afresh. Also, xorm requires deletes to have a
// condition, and we want to delete everything, thus 1=1.
if err := db.DeleteAllRecords("repo_indexer_status"); err != nil {
log.Fatal("System error: %v", err)
}
var maxRepoID int64
if maxRepoID, err = db.GetMaxID("repository"); err != nil {
log.Fatal("System error: %v", err)
}
// start with the maximum existing repo ID and work backwards, so that we
// don't include repos that are created after gitea starts; such repos will
// already be added to the indexer, and we don't need to add them again.
for maxRepoID > 0 {
select {
case <-ctx.Done():
log.Info("Repository Indexer population shutdown before completion")
return
default:
}
ids, err := repo_model.GetUnindexedRepos(repo_model.RepoIndexerTypeCode, maxRepoID, 0, 50)
if err != nil {
log.Error("populateRepoIndexer: %v", err)
return
} else if len(ids) == 0 {
break
}
for _, id := range ids {
select {
case <-ctx.Done():
log.Info("Repository Indexer population shutdown before completion")
return
default:
}
if err := indexerQueue.Push(&internal.IndexerData{RepoID: id}); err != nil {
log.Error("indexerQueue.Push: %v", err)
return
}
maxRepoID = id - 1
}
}
log.Info("Done (re)populating the repo indexer with existing repositories")
}