You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.

186 lines
5.2 KiB

3 years ago
3 years ago
3 years ago
3 years ago
3 years ago
3 years ago
3 years ago
3 years ago
3 years ago
3 years ago
Fix dead lock (#5815) * reduce locks to avoid dead lock Flush->FlushData->uplloadPipeline.FluahAll uploaderCount>0 goroutine 1 [sync.Cond.Wait, 71 minutes]: sync.runtime_notifyListWait(0xc0007ae4d0, 0x0) /usr/local/go/src/runtime/sema.go:569 +0x159 sync.(*Cond).Wait(0xc001a59290?) /usr/local/go/src/sync/cond.go:70 +0x85 github.com/seaweedfs/seaweedfs/weed/mount/page_writer.(*UploadPipeline).waitForCurrentWritersToComplete(0xc0002ee4d0) /github/workspace/weed/mount/page_writer/upload_pipeline_lock.go:58 +0x32 github.com/seaweedfs/seaweedfs/weed/mount/page_writer.(*UploadPipeline).FlushAll(0xc0002ee4d0) /github/workspace/weed/mount/page_writer/upload_pipeline.go:151 +0x25 github.com/seaweedfs/seaweedfs/weed/mount.(*ChunkedDirtyPages).FlushData(0xc00087e840) /github/workspace/weed/mount/dirty_pages_chunked.go:54 +0x29 github.com/seaweedfs/seaweedfs/weed/mount.(*PageWriter).FlushData(...) /github/workspace/weed/mount/page_writer.go:50 github.com/seaweedfs/seaweedfs/weed/mount.(*WFS).doFlush(0xc0006ad600, 0xc00030d380, 0x0, 0x0) /github/workspace/weed/mount/weedfs_file_sync.go:101 +0x169 github.com/seaweedfs/seaweedfs/weed/mount.(*WFS).Flush(0xc0006ad600, 0xc001a594a8?, 0xc0004c1ca0) /github/workspace/weed/mount/weedfs_file_sync.go:59 +0x48 github.com/hanwen/go-fuse/v2/fuse.doFlush(0xc0000da870?, 0xc0004c1b08) SaveContent -> MemChunk.RLock -> ChunkedDirtyPages.saveChunkedFileIntervalToStorage pages.fh.AddChunks([]*filer_pb.FileChunk{chunk}) fh.entryLock.Lock() sync.(*RWMutex).Lock(0x0?) /usr/local/go/src/sync/rwmutex.go:146 +0x31 github.com/seaweedfs/seaweedfs/weed/mount.(*FileHandle).AddChunks(0xc00030d380, {0xc00028bdc8, 0x1, 0x1}) /github/workspace/weed/mount/filehandle.go:93 +0x45 github.com/seaweedfs/seaweedfs/weed/mount.(*ChunkedDirtyPages).saveChunkedFileIntervalToStorage(0xc00087e840, {0x2be7ac0, 0xc00018d9e0}, 0x0, 0x121, 0x17e3c624565ace45, 0x1?) /github/workspace/weed/mount/dirty_pages_chunked.go:80 +0x2d4 github.com/seaweedfs/seaweedfs/weed/mount/page_writer.(*MemChunk).SaveContent(0xc0008d9130, 0xc0008093e0) /github/workspace/weed/mount/page_writer/page_chunk_mem.go:115 +0x112 github.com/seaweedfs/seaweedfs/weed/mount/page_writer.(*UploadPipeline).moveToSealed.func1() /github/workspace/weed/mount/page_writer/upload_pipeline.go:187 +0x55 github.com/seaweedfs/seaweedfs/weed/util.(*LimitedConcurrentExecutor).Execute.func1() /github/workspace/weed/util/limited_executor.go:38 +0x62 created by github.com/seaweedfs/seaweedfs/weed/util.(*LimitedConcurrentExecutor).Execute in goroutine 1 /github/workspace/weed/util/limited_executor.go:33 +0x97 On metadata update fh.entryLock.Lock() fh.dirtyPages.Destroy() up.chunksLock.Lock => each sealed chunk.FreeReference => MemChunk.Lock goroutine 134 [sync.RWMutex.Lock, 71 minutes]: sync.runtime_SemacquireRWMutex(0xc0007c3558?, 0xea?, 0x3fb0800?) /usr/local/go/src/runtime/sema.go:87 +0x25 sync.(*RWMutex).Lock(0xc0007c35a8?) /usr/local/go/src/sync/rwmutex.go:151 +0x6a github.com/seaweedfs/seaweedfs/weed/mount/page_writer.(*MemChunk).FreeResource(0xc0008d9130) /github/workspace/weed/mount/page_writer/page_chunk_mem.go:38 +0x2a github.com/seaweedfs/seaweedfs/weed/mount/page_writer.(*SealedChunk).FreeReference(0xc00071cdb0, {0xc0006ba1a0, 0x20}) /github/workspace/weed/mount/page_writer/upload_pipeline.go:38 +0xb7 github.com/seaweedfs/seaweedfs/weed/mount/page_writer.(*UploadPipeline).Shutdown(0xc0002ee4d0) /github/workspace/weed/mount/page_writer/upload_pipeline.go:220 +0x185 github.com/seaweedfs/seaweedfs/weed/mount.(*ChunkedDirtyPages).Destroy(0xc0008cea40?) /github/workspace/weed/mount/dirty_pages_chunked.go:87 +0x17 github.com/seaweedfs/seaweedfs/weed/mount.(*PageWriter).Destroy(...) /github/workspace/weed/mount/page_writer.go:78 github.com/seaweedfs/seaweedfs/weed/mount.NewSeaweedFileSystem.func3({0xc00069a6c0, 0x30}, 0x6?) /github/workspace/weed/mount/weedfs.go:119 +0x17a github.com/seaweedfs/seaweedfs/weed/mount/meta_cache.NewMetaCache.func1({0xc00069a6c0?, 0xc00069a480?}, 0x4015b40?) /github/workspace/weed/mount/meta_cache/meta_cache.go:37 +0x1c github.com/seaweedfs/seaweedfs/weed/mount/meta_cache.SubscribeMetaEvents.func1(0xc000661810) /github/workspace/weed/mount/meta_cache/meta_cache_subscribe.go:43 +0x570 * use locked entry everywhere * modifiable remote entry * skip locking after getting lock from fhLockTable
5 months ago
3 years ago
more solid weed mount (#4089) * compare chunks by timestamp * fix slab clearing error * fix test compilation * move oldest chunk to sealed, instead of by fullness * lock on fh.entryViewCache * remove verbose logs * revert slat clearing * less logs * less logs * track write and read by timestamp * remove useless logic * add entry lock on file handle release * use mem chunk only, swap file chunk has problems * comment out code that maybe used later * add debug mode to compare data read and write * more efficient readResolvedChunks with linked list * small optimization * fix test compilation * minor fix on writer * add SeparateGarbageChunks * group chunks into sections * turn off debug mode * fix tests * fix tests * tmp enable swap file chunk * Revert "tmp enable swap file chunk" This reverts commit 985137ec472924e4815f258189f6ca9f2168a0a7. * simple refactoring * simple refactoring * do not re-use swap file chunk. Sealed chunks should not be re-used. * comment out debugging facilities * either mem chunk or swap file chunk is fine now * remove orderedMutex as *semaphore.Weighted not found impactful * optimize size calculation for changing large files * optimize performance to avoid going through the long list of chunks * still problems with swap file chunk * rename * tiny optimization * swap file chunk save only successfully read data * fix * enable both mem and swap file chunk * resolve chunks with range * rename * fix chunk interval list * also change file handle chunk group when adding chunks * pick in-active chunk with time-decayed counter * fix compilation * avoid nil with empty fh.entry * refactoring * rename * rename * refactor visible intervals to *list.List * refactor chunkViews to *list.List * add IntervalList for generic interval list * change visible interval to use IntervalList in generics * cahnge chunkViews to *IntervalList[*ChunkView] * use NewFileChunkSection to create * rename variables * refactor * fix renaming leftover * renaming * renaming * add insert interval * interval list adds lock * incrementally add chunks to readers Fixes: 1. set start and stop offset for the value object 2. clone the value object 3. use pointer instead of copy-by-value when passing to interval.Value 4. use insert interval since adding chunk could be out of order * fix tests compilation * fix tests compilation
2 years ago
3 years ago
3 years ago
3 years ago
more solid weed mount (#4089) * compare chunks by timestamp * fix slab clearing error * fix test compilation * move oldest chunk to sealed, instead of by fullness * lock on fh.entryViewCache * remove verbose logs * revert slat clearing * less logs * less logs * track write and read by timestamp * remove useless logic * add entry lock on file handle release * use mem chunk only, swap file chunk has problems * comment out code that maybe used later * add debug mode to compare data read and write * more efficient readResolvedChunks with linked list * small optimization * fix test compilation * minor fix on writer * add SeparateGarbageChunks * group chunks into sections * turn off debug mode * fix tests * fix tests * tmp enable swap file chunk * Revert "tmp enable swap file chunk" This reverts commit 985137ec472924e4815f258189f6ca9f2168a0a7. * simple refactoring * simple refactoring * do not re-use swap file chunk. Sealed chunks should not be re-used. * comment out debugging facilities * either mem chunk or swap file chunk is fine now * remove orderedMutex as *semaphore.Weighted not found impactful * optimize size calculation for changing large files * optimize performance to avoid going through the long list of chunks * still problems with swap file chunk * rename * tiny optimization * swap file chunk save only successfully read data * fix * enable both mem and swap file chunk * resolve chunks with range * rename * fix chunk interval list * also change file handle chunk group when adding chunks * pick in-active chunk with time-decayed counter * fix compilation * avoid nil with empty fh.entry * refactoring * rename * rename * refactor visible intervals to *list.List * refactor chunkViews to *list.List * add IntervalList for generic interval list * change visible interval to use IntervalList in generics * cahnge chunkViews to *IntervalList[*ChunkView] * use NewFileChunkSection to create * rename variables * refactor * fix renaming leftover * renaming * renaming * add insert interval * interval list adds lock * incrementally add chunks to readers Fixes: 1. set start and stop offset for the value object 2. clone the value object 3. use pointer instead of copy-by-value when passing to interval.Value 4. use insert interval since adding chunk could be out of order * fix tests compilation * fix tests compilation
2 years ago
3 years ago
  1. package mount
  2. import (
  3. "context"
  4. "fmt"
  5. "github.com/hanwen/go-fuse/v2/fuse"
  6. "github.com/seaweedfs/seaweedfs/weed/filer"
  7. "github.com/seaweedfs/seaweedfs/weed/glog"
  8. "github.com/seaweedfs/seaweedfs/weed/pb/filer_pb"
  9. "github.com/seaweedfs/seaweedfs/weed/util"
  10. "syscall"
  11. "time"
  12. )
  13. /**
  14. * Flush method
  15. *
  16. * This is called on each close() of the opened file.
  17. *
  18. * Since file descriptors can be duplicated (dup, dup2, fork), for
  19. * one open call there may be many flush calls.
  20. *
  21. * Filesystems shouldn't assume that flush will always be called
  22. * after some writes, or that if will be called at all.
  23. *
  24. * fi->fh will contain the value set by the open method, or will
  25. * be undefined if the open method didn't set any value.
  26. *
  27. * NOTE: the name of the method is misleading, since (unlike
  28. * fsync) the filesystem is not forced to flush pending writes.
  29. * One reason to flush data is if the filesystem wants to return
  30. * write errors during close. However, such use is non-portable
  31. * because POSIX does not require [close] to wait for delayed I/O to
  32. * complete.
  33. *
  34. * If the filesystem supports file locking operations (setlk,
  35. * getlk) it should remove all locks belonging to 'fi->owner'.
  36. *
  37. * If this request is answered with an error code of ENOSYS,
  38. * this is treated as success and future calls to flush() will
  39. * succeed automatically without being send to the filesystem
  40. * process.
  41. *
  42. * Valid replies:
  43. * fuse_reply_err
  44. *
  45. * @param req request handle
  46. * @param ino the inode number
  47. * @param fi file information
  48. *
  49. * [close]: http://pubs.opengroup.org/onlinepubs/9699919799/functions/close.html
  50. */
  51. func (wfs *WFS) Flush(cancel <-chan struct{}, in *fuse.FlushIn) fuse.Status {
  52. fh := wfs.GetHandle(FileHandleId(in.Fh))
  53. if fh == nil {
  54. return fuse.ENOENT
  55. }
  56. return wfs.doFlush(fh, in.Uid, in.Gid)
  57. }
  58. /**
  59. * Synchronize file contents
  60. *
  61. * If the datasync parameter is non-zero, then only the user data
  62. * should be flushed, not the meta data.
  63. *
  64. * If this request is answered with an error code of ENOSYS,
  65. * this is treated as success and future calls to fsync() will
  66. * succeed automatically without being send to the filesystem
  67. * process.
  68. *
  69. * Valid replies:
  70. * fuse_reply_err
  71. *
  72. * @param req request handle
  73. * @param ino the inode number
  74. * @param datasync flag indicating if only data should be flushed
  75. * @param fi file information
  76. */
  77. func (wfs *WFS) Fsync(cancel <-chan struct{}, in *fuse.FsyncIn) (code fuse.Status) {
  78. fh := wfs.GetHandle(FileHandleId(in.Fh))
  79. if fh == nil {
  80. return fuse.ENOENT
  81. }
  82. return wfs.doFlush(fh, in.Uid, in.Gid)
  83. }
  84. func (wfs *WFS) doFlush(fh *FileHandle, uid, gid uint32) fuse.Status {
  85. // flush works at fh level
  86. fileFullPath := fh.FullPath()
  87. dir, name := fileFullPath.DirAndName()
  88. // send the data to the OS
  89. glog.V(4).Infof("doFlush %s fh %d", fileFullPath, fh.fh)
  90. if !wfs.IsOverQuota {
  91. if err := fh.dirtyPages.FlushData(); err != nil {
  92. glog.Errorf("%v doFlush: %v", fileFullPath, err)
  93. return fuse.EIO
  94. }
  95. }
  96. if !fh.dirtyMetadata {
  97. return fuse.OK
  98. }
  99. if wfs.IsOverQuota {
  100. return fuse.Status(syscall.ENOSPC)
  101. }
  102. fhActiveLock := fh.wfs.fhLockTable.AcquireLock("doFlush", fh.fh, util.ExclusiveLock)
  103. defer fh.wfs.fhLockTable.ReleaseLock(fh.fh, fhActiveLock)
  104. err := wfs.WithFilerClient(false, func(client filer_pb.SeaweedFilerClient) error {
  105. entry := fh.GetEntry()
  106. entry.Name = name // this flush may be just after a rename operation
  107. if entry.Attributes != nil {
  108. entry.Attributes.Mime = fh.contentType
  109. if entry.Attributes.Uid == 0 {
  110. entry.Attributes.Uid = uid
  111. }
  112. if entry.Attributes.Gid == 0 {
  113. entry.Attributes.Gid = gid
  114. }
  115. if entry.Attributes.Crtime == 0 {
  116. entry.Attributes.Crtime = time.Now().Unix()
  117. }
  118. entry.Attributes.Mtime = time.Now().Unix()
  119. }
  120. request := &filer_pb.CreateEntryRequest{
  121. Directory: string(dir),
  122. Entry: entry.GetEntry(),
  123. Signatures: []int32{wfs.signature},
  124. SkipCheckParentDirectory: true,
  125. }
  126. glog.V(4).Infof("%s set chunks: %v", fileFullPath, len(entry.GetChunks()))
  127. //for i, chunk := range entry.GetChunks() {
  128. // glog.V(4).Infof("%s chunks %d: %v [%d,%d)", fileFullPath, i, chunk.GetFileIdString(), chunk.Offset, chunk.Offset+int64(chunk.Size))
  129. //}
  130. manifestChunks, nonManifestChunks := filer.SeparateManifestChunks(entry.GetChunks())
  131. chunks, _ := filer.CompactFileChunks(wfs.LookupFn(), nonManifestChunks)
  132. chunks, manifestErr := filer.MaybeManifestize(wfs.saveDataAsChunk(fileFullPath), chunks)
  133. if manifestErr != nil {
  134. // not good, but should be ok
  135. glog.V(0).Infof("MaybeManifestize: %v", manifestErr)
  136. }
  137. entry.Chunks = append(chunks, manifestChunks...)
  138. wfs.mapPbIdFromLocalToFiler(request.Entry)
  139. defer wfs.mapPbIdFromFilerToLocal(request.Entry)
  140. if err := filer_pb.CreateEntry(client, request); err != nil {
  141. glog.Errorf("fh flush create %s: %v", fileFullPath, err)
  142. return fmt.Errorf("fh flush create %s: %v", fileFullPath, err)
  143. }
  144. wfs.metaCache.InsertEntry(context.Background(), filer.FromPbEntry(request.Directory, request.Entry))
  145. return nil
  146. })
  147. if err == nil {
  148. fh.dirtyMetadata = false
  149. }
  150. if err != nil {
  151. glog.Errorf("%v fh %d flush: %v", fileFullPath, fh.fh, err)
  152. return fuse.EIO
  153. }
  154. if IsDebugFileReadWrite {
  155. fh.mirrorFile.Sync()
  156. }
  157. return fuse.OK
  158. }