broker_topic_partition_read_write.go 5.7 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179
  1. package broker
  2. import (
  3. "fmt"
  4. "github.com/seaweedfs/seaweedfs/weed/filer"
  5. "github.com/seaweedfs/seaweedfs/weed/glog"
  6. "github.com/seaweedfs/seaweedfs/weed/mq/topic"
  7. "github.com/seaweedfs/seaweedfs/weed/pb/filer_pb"
  8. "github.com/seaweedfs/seaweedfs/weed/pb/mq_pb"
  9. "github.com/seaweedfs/seaweedfs/weed/util"
  10. "github.com/seaweedfs/seaweedfs/weed/util/log_buffer"
  11. "google.golang.org/protobuf/proto"
  12. "math"
  13. "sync/atomic"
  14. "time"
  15. util_http "github.com/seaweedfs/seaweedfs/weed/util/http"
  16. )
  17. func (b *MessageQueueBroker) genLogFlushFunc(t topic.Topic, partition *mq_pb.Partition) log_buffer.LogFlushFuncType {
  18. topicDir := fmt.Sprintf("%s/%s/%s", filer.TopicsDir, t.Namespace, t.Name)
  19. partitionGeneration := time.Unix(0, partition.UnixTimeNs).UTC().Format(topic.TIME_FORMAT)
  20. partitionDir := fmt.Sprintf("%s/%s/%04d-%04d", topicDir, partitionGeneration, partition.RangeStart, partition.RangeStop)
  21. return func(logBuffer *log_buffer.LogBuffer, startTime, stopTime time.Time, buf []byte) {
  22. if len(buf) == 0 {
  23. return
  24. }
  25. startTime, stopTime = startTime.UTC(), stopTime.UTC()
  26. targetFile := fmt.Sprintf("%s/%s", partitionDir, startTime.Format(topic.TIME_FORMAT))
  27. // TODO append block with more metadata
  28. for {
  29. if err := b.appendToFile(targetFile, buf); err != nil {
  30. glog.V(0).Infof("metadata log write failed %s: %v", targetFile, err)
  31. time.Sleep(737 * time.Millisecond)
  32. } else {
  33. break
  34. }
  35. }
  36. atomic.StoreInt64(&logBuffer.LastFlushTsNs, stopTime.UnixNano())
  37. b.accessLock.Lock()
  38. defer b.accessLock.Unlock()
  39. p := topic.FromPbPartition(partition)
  40. if localPartition := b.localTopicManager.GetLocalPartition(t, p); localPartition != nil {
  41. localPartition.NotifyLogFlushed(logBuffer.LastFlushTsNs)
  42. }
  43. println("flushing at", logBuffer.LastFlushTsNs, "to", targetFile, "size", len(buf))
  44. }
  45. }
  46. func (b *MessageQueueBroker) genLogOnDiskReadFunc(t topic.Topic, partition *mq_pb.Partition) log_buffer.LogReadFromDiskFuncType {
  47. topicDir := fmt.Sprintf("%s/%s/%s", filer.TopicsDir, t.Namespace, t.Name)
  48. partitionGeneration := time.Unix(0, partition.UnixTimeNs).UTC().Format(topic.TIME_FORMAT)
  49. partitionDir := fmt.Sprintf("%s/%s/%04d-%04d", topicDir, partitionGeneration, partition.RangeStart, partition.RangeStop)
  50. lookupFileIdFn := func(fileId string) (targetUrls []string, err error) {
  51. return b.MasterClient.LookupFileId(fileId)
  52. }
  53. eachChunkFn := func(buf []byte, eachLogEntryFn log_buffer.EachLogEntryFuncType, starTsNs, stopTsNs int64) (processedTsNs int64, err error) {
  54. for pos := 0; pos+4 < len(buf); {
  55. size := util.BytesToUint32(buf[pos : pos+4])
  56. if pos+4+int(size) > len(buf) {
  57. err = fmt.Errorf("LogOnDiskReadFunc: read [%d,%d) from [0,%d)", pos, pos+int(size)+4, len(buf))
  58. return
  59. }
  60. entryData := buf[pos+4 : pos+4+int(size)]
  61. logEntry := &filer_pb.LogEntry{}
  62. if err = proto.Unmarshal(entryData, logEntry); err != nil {
  63. pos += 4 + int(size)
  64. err = fmt.Errorf("unexpected unmarshal mq_pb.Message: %v", err)
  65. return
  66. }
  67. if logEntry.TsNs < starTsNs {
  68. pos += 4 + int(size)
  69. continue
  70. }
  71. if stopTsNs != 0 && logEntry.TsNs > stopTsNs {
  72. println("stopTsNs", stopTsNs, "logEntry.TsNs", logEntry.TsNs)
  73. return
  74. }
  75. if _, err = eachLogEntryFn(logEntry); err != nil {
  76. err = fmt.Errorf("process log entry %v: %v", logEntry, err)
  77. return
  78. }
  79. processedTsNs = logEntry.TsNs
  80. pos += 4 + int(size)
  81. }
  82. return
  83. }
  84. eachFileFn := func(entry *filer_pb.Entry, eachLogEntryFn log_buffer.EachLogEntryFuncType, starTsNs, stopTsNs int64) (processedTsNs int64, err error) {
  85. if len(entry.Content) > 0 {
  86. glog.Warningf("this should not happen. unexpected content in %s/%s", partitionDir, entry.Name)
  87. return
  88. }
  89. var urlStrings []string
  90. for _, chunk := range entry.Chunks {
  91. if chunk.Size == 0 {
  92. continue
  93. }
  94. if chunk.IsChunkManifest {
  95. glog.Warningf("this should not happen. unexpected chunk manifest in %s/%s", partitionDir, entry.Name)
  96. return
  97. }
  98. urlStrings, err = lookupFileIdFn(chunk.FileId)
  99. if err != nil {
  100. err = fmt.Errorf("lookup %s: %v", chunk.FileId, err)
  101. return
  102. }
  103. if len(urlStrings) == 0 {
  104. err = fmt.Errorf("no url found for %s", chunk.FileId)
  105. return
  106. }
  107. // try one of the urlString until util.Get(urlString) succeeds
  108. var processed bool
  109. for _, urlString := range urlStrings {
  110. // TODO optimization opportunity: reuse the buffer
  111. var data []byte
  112. if data, _, err = util_http.Get(urlString); err == nil {
  113. processed = true
  114. if processedTsNs, err = eachChunkFn(data, eachLogEntryFn, starTsNs, stopTsNs); err != nil {
  115. return
  116. }
  117. break
  118. }
  119. }
  120. if !processed {
  121. err = fmt.Errorf("no data processed for %s %s", entry.Name, chunk.FileId)
  122. return
  123. }
  124. }
  125. return
  126. }
  127. return func(startPosition log_buffer.MessagePosition, stopTsNs int64, eachLogEntryFn log_buffer.EachLogEntryFuncType) (lastReadPosition log_buffer.MessagePosition, isDone bool, err error) {
  128. startFileName := startPosition.UTC().Format(topic.TIME_FORMAT)
  129. startTsNs := startPosition.Time.UnixNano()
  130. stopTime := time.Unix(0, stopTsNs)
  131. var processedTsNs int64
  132. err = b.WithFilerClient(false, func(client filer_pb.SeaweedFilerClient) error {
  133. return filer_pb.SeaweedList(client, partitionDir, "", func(entry *filer_pb.Entry, isLast bool) error {
  134. if entry.IsDirectory {
  135. return nil
  136. }
  137. if stopTsNs != 0 && entry.Name > stopTime.UTC().Format(topic.TIME_FORMAT) {
  138. isDone = true
  139. return nil
  140. }
  141. if entry.Name < startPosition.UTC().Format(topic.TIME_FORMAT) {
  142. return nil
  143. }
  144. if processedTsNs, err = eachFileFn(entry, eachLogEntryFn, startTsNs, stopTsNs); err != nil {
  145. return err
  146. }
  147. return nil
  148. }, startFileName, true, math.MaxInt32)
  149. })
  150. lastReadPosition = log_buffer.NewMessagePosition(processedTsNs, -2)
  151. return
  152. }
  153. }