|
|
package main
import ( "flag" "fmt" "io" "os" "path" "strconv"
"github.com/chrislusf/seaweedfs/weed/glog" "github.com/chrislusf/seaweedfs/weed/storage" "github.com/chrislusf/seaweedfs/weed/util" )
var ( fixVolumePath = flag.String("dir", "/tmp", "data directory to store files") fixVolumeCollection = flag.String("collection", "", "the volume collection name") fixVolumeId = flag.Int("volumeId", -1, "a volume id. The volume should already exist in the dir. The volume index file should not exist.") )
/* This is to resolve an one-time issue that caused inconsistency with .dat and .idx files. In this case, the .dat file contains all data, but some of deletion caused incorrect offset. The .idx has all correct offsets.
1. fix the .dat file, a new .dat_fixed file will be generated. go run fix_dat.go -volumeId=9 -dir=/Users/chrislu/Downloads 2. move the original .dat and .idx files to some backup folder, and rename .dat_fixed to .dat file mv 9.dat_fixed 9.dat 3. fix the .idx file with the "weed fix" weed fix -volumeId=9 -dir=/Users/chrislu/Downloads */ func main() { flag.Parse() fileName := strconv.Itoa(*fixVolumeId) if *fixVolumeCollection != "" { fileName = *fixVolumeCollection + "_" + fileName } indexFile, err := os.OpenFile(path.Join(*fixVolumePath, fileName+".idx"), os.O_RDONLY, 0644) if err != nil { glog.Fatalf("Read Volume Index [ERROR] %s\n", err) } defer indexFile.Close() datFile, err := os.OpenFile(path.Join(*fixVolumePath, fileName+".dat"), os.O_RDONLY, 0644) if err != nil { glog.Fatalf("Read Volume Data [ERROR] %s\n", err) } defer datFile.Close()
newDatFile, err := os.Create(path.Join(*fixVolumePath, fileName+".dat_fixed")) if err != nil { glog.Fatalf("Write New Volume Data [ERROR] %s\n", err) } defer newDatFile.Close()
header := make([]byte, storage.SuperBlockSize) datFile.Read(header) newDatFile.Write(header)
iterateEntries(datFile, indexFile, func(n *storage.Needle, offset int64) { fmt.Printf("file id=%d name=%s size=%d dataSize=%d\n", n.Id, string(n.Name), n.Size, n.DataSize) s, e := n.Append(newDatFile, storage.Version2) fmt.Printf("size %d error %v\n", s, e) })
}
func iterateEntries(datFile, idxFile *os.File, visitNeedle func(n *storage.Needle, offset int64)) { // start to read index file
var readerOffset int64 bytes := make([]byte, 16) count, _ := idxFile.ReadAt(bytes, readerOffset) readerOffset += int64(count)
// start to read dat file
offset := int64(storage.SuperBlockSize) version := storage.Version2 n, rest, err := storage.ReadNeedleHeader(datFile, version, offset) if err != nil { fmt.Printf("cannot read needle header: %v", err) return } fmt.Printf("Needle %+v, rest %d\n", n, rest) for n != nil && count > 0 { // parse index file entry
key := util.BytesToUint64(bytes[0:8]) offsetFromIndex := util.BytesToUint32(bytes[8:12]) sizeFromIndex := util.BytesToUint32(bytes[12:16]) count, _ = idxFile.ReadAt(bytes, readerOffset) readerOffset += int64(count)
if offsetFromIndex != 0 && offset != int64(offsetFromIndex)*8 { //t := offset
offset = int64(offsetFromIndex) * 8 //fmt.Printf("Offset change %d => %d\n", t, offset)
}
fmt.Printf("key: %d offsetFromIndex %d n.Size %d sizeFromIndex:%d\n", key, offsetFromIndex, n.Size, sizeFromIndex)
padding := storage.NeedlePaddingSize - ((sizeFromIndex + storage.NeedleHeaderSize + storage.NeedleChecksumSize) % storage.NeedlePaddingSize) rest = sizeFromIndex + storage.NeedleChecksumSize + padding
func() { defer func() { if r := recover(); r != nil { fmt.Println("Recovered in f", r) } }() if err = n.ReadNeedleBody(datFile, version, offset+int64(storage.NeedleHeaderSize), rest); err != nil { fmt.Printf("cannot read needle body: offset %d body %d %v\n", offset, rest, err) } }()
if n.Size <= n.DataSize { continue } visitNeedle(n, offset)
offset += int64(storage.NeedleHeaderSize) + int64(rest) //fmt.Printf("==> new entry offset %d\n", offset)
if n, rest, err = storage.ReadNeedleHeader(datFile, version, offset); err != nil { if err == io.EOF { return }
fmt.Printf("cannot read needle header: %v\n", err) return } //fmt.Printf("new entry needle size:%d rest:%d\n", n.Size, rest)
}
}
|