1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191
|
package streamcache
import (
"context"
"crypto/rand"
"fmt"
"io"
"os"
"path/filepath"
"sync"
"time"
"github.com/prometheus/client_golang/prometheus"
"github.com/prometheus/client_golang/prometheus/promauto"
"github.com/sirupsen/logrus"
"gitlab.com/gitlab-org/gitaly/v16/internal/dontpanic"
"gitlab.com/gitlab-org/gitaly/v16/internal/helper/perm"
)
var (
fileRemoveCounter = promauto.NewCounterVec(
prometheus.CounterOpts{
Name: "gitaly_streamcache_filestore_removed_total",
Help: "Number of files removed from streamcache via file walks",
},
[]string{"dir"},
)
diskUsageGauge = promauto.NewGaugeVec(
prometheus.GaugeOpts{
Name: "gitaly_streamcache_filestore_disk_usage_bytes",
Help: "Disk usage per filestore",
},
[]string{"dir"},
)
)
// Filestore creates temporary files in dir. These files get deleted once
// they are older (by mtime) than maxAge via a goroutine that does
// periodic file walks. To make these file walks efficient, we spread the
// temporary files across a balanced directory tree in subdirectories
// aa/my-temp-file where aa is a hexadecimal number. This means
// we eventually create exactly 256 subdirectories.
//
// Each instance of filestore carries a random token which is used as a
// prefix in the tempfiles it creates. This ensures that it is safe to
// have multiple instances of filestore use the same directory on disk;
// their tempfiles will not collide. There is one caveat: if multiple
// filestores share the same directory but have different maximum ages,
// then lowest maximum age becomes the effective maximum age for all of
// them.
type filestore struct {
dir string
maxAge time.Duration
m sync.Mutex
id []byte
counter uint64
stop chan struct{}
sleepLoop *dontpanic.Forever
}
func newFilestore(dir string, maxAge time.Duration, sleep func(time.Duration) <-chan time.Time, logger logrus.FieldLogger) *filestore {
fs := &filestore{
dir: dir,
maxAge: maxAge,
stop: make(chan struct{}),
sleepLoop: dontpanic.NewForever(time.Minute),
}
fs.sleepLoop.Go(func() {
sleepLoop(fs.stop, fs.maxAge, sleep, func() {
diskUsageGauge.WithLabelValues(fs.dir).Set(fs.diskUsage())
if err := fs.cleanWalk(time.Now().Add(-fs.maxAge)); err != nil {
logger.WithError(err).Error("streamcache filestore cleanup")
}
})
})
return fs
}
type namedWriteCloser interface {
Name() string
io.WriteCloser
}
// Create creates a new tempfile. It does not use os.CreateTemp because
// the documentation of TempFile makes no promises about reusing tempfile
// names after a file has been deleted. By using a very large (uint64)
// counter, Create makes it clear / explicit how unlikely reuse is.
func (fs *filestore) Create() (namedWriteCloser, error) {
if err := fs.ensureCacheID(); err != nil {
return nil, err
}
fileID := fs.nextFileID()
name := fmt.Sprintf("%x-%d",
// fs.id ensures uniqueness among other *filestore instances
fs.id,
// fileID ensures uniqueness (modulo roll-over) among other files
// created by this *filestore instance
fileID,
)
path := filepath.Join(fs.dir, fmt.Sprintf("%02x", uint8(fileID)), name)
if err := os.MkdirAll(filepath.Dir(path), perm.PrivateDir); err != nil {
return nil, fmt.Errorf("Create: mkdir: %w", err)
}
f, err := os.OpenFile(path, os.O_WRONLY|os.O_CREATE|os.O_EXCL, perm.SharedFile)
if err != nil {
return nil, fmt.Errorf("Create: %w", err)
}
return f, nil
}
func (fs *filestore) ensureCacheID() error {
fs.m.Lock()
defer fs.m.Unlock()
if len(fs.id) == 0 {
buf := make([]byte, 10)
if _, err := io.ReadFull(rand.Reader, buf); err != nil {
return err
}
fs.id = buf
}
return nil
}
func (fs *filestore) nextFileID() uint64 {
fs.m.Lock()
defer fs.m.Unlock()
fs.counter++
return fs.counter
}
func (fs *filestore) Stop() {
fs.m.Lock()
defer fs.m.Unlock()
select {
case <-fs.stop:
default:
close(fs.stop)
}
fs.sleepLoop.Cancel()
}
// cleanWalk removes files but not directories. This is to avoid races
// when a directory looks empty but another goroutine is about to create
// a new file in it with fs.Create(). Because the number of directories
// is bounded by the directory scheme to 256, there is no need to remove
// the directories anyway.
func (fs *filestore) cleanWalk(cutoff time.Time) error {
// If a server reset has left some directories in a bad state, this will
// fix it.
if err := perm.FixDirectoryPermissions(context.Background(), fs.dir); err != nil {
return err
}
return filepath.Walk(fs.dir, func(path string, info os.FileInfo, err error) error {
if err == nil && !info.IsDir() && info.ModTime().Before(cutoff) {
err = os.Remove(path)
fileRemoveCounter.WithLabelValues(fs.dir).Inc()
}
if os.IsNotExist(err) {
err = nil
}
return err
})
}
func (fs *filestore) diskUsage() float64 {
var total float64
_ = filepath.Walk(fs.dir, func(_ string, info os.FileInfo, _ error) error {
if info != nil {
total += float64(info.Size())
}
return nil
})
return total
}
|