-
Notifications
You must be signed in to change notification settings - Fork 9.7k
Commit
This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository.
Utilize WAL to patch operation history
Signed-off-by: Marek Siarkowicz <siarkowicz@google.com>
- Loading branch information
Showing
8 changed files
with
296 additions
and
459 deletions.
There are no files selected for viewing
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1,188 @@ | ||
| package report | ||
|
|
||
| import ( | ||
| "errors" | ||
| "fmt" | ||
| "github.com/google/go-cmp/cmp" | ||
| pb "go.etcd.io/etcd/api/v3/etcdserverpb" | ||
| "go.etcd.io/etcd/server/v3/storage/wal" | ||
| "go.etcd.io/etcd/server/v3/storage/wal/walpb" | ||
| "go.etcd.io/etcd/tests/v3/framework/e2e" | ||
| "go.etcd.io/etcd/tests/v3/robustness/model" | ||
| "go.etcd.io/raft/v3/raftpb" | ||
| "go.uber.org/zap" | ||
| "io" | ||
| "os" | ||
| "path/filepath" | ||
| "strings" | ||
| ) | ||
|
|
||
| func LoadClusterPersistedRequests(lg *zap.Logger, path string) ([]model.EtcdRequest, error) { | ||
| files, err := os.ReadDir(path) | ||
| if err != nil { | ||
| return nil, err | ||
| } | ||
| dataDirs := []string{} | ||
| for _, file := range files { | ||
| if file.IsDir() && strings.HasPrefix(file.Name(), "server-") { | ||
| dataDirs = append(dataDirs, filepath.Join(path, file.Name())) | ||
| } | ||
| } | ||
| return PersistedRequestsDirs(lg, dataDirs) | ||
| } | ||
|
|
||
| func PersistedRequestsCluster(lg *zap.Logger, cluster *e2e.EtcdProcessCluster) ([]model.EtcdRequest, error) { | ||
| dataDirs := []string{} | ||
| for _, proc := range cluster.Procs { | ||
| dataDirs = append(dataDirs, proc.Config().DataDirPath) | ||
| } | ||
| return PersistedRequestsDirs(lg, dataDirs) | ||
| } | ||
|
|
||
| func PersistedRequestsDirs(lg *zap.Logger, dataDirs []string) ([]model.EtcdRequest, error) { | ||
| persistedRequests := []model.EtcdRequest{} | ||
| for _, dir := range dataDirs { | ||
| memberRequests, err := requestsPersistedInWAL(lg, dir) | ||
| if err != nil { | ||
| panic(err) | ||
| } | ||
| minLength := min(len(persistedRequests), len(memberRequests)) | ||
| if diff := cmp.Diff(memberRequests[:minLength], persistedRequests[:minLength]); diff != "" { | ||
| return nil, fmt.Errorf("unexpected differences between wal entries, diff:\n%s", diff) | ||
| } | ||
| if len(memberRequests) > len(persistedRequests) { | ||
| persistedRequests = memberRequests | ||
| } | ||
| } | ||
| return persistedRequests, nil | ||
| } | ||
|
|
||
| func requestsPersistedInWAL(lg *zap.Logger, dataDir string) ([]model.EtcdRequest, error) { | ||
| state, ents, err := ReadWAL(lg, dataDir) | ||
| if err != nil { | ||
| return nil, err | ||
| } | ||
| requests := make([]model.EtcdRequest, 0, len(ents)) | ||
| for _, ent := range ents { | ||
| if ent.Type != raftpb.EntryNormal || len(ent.Data) == 0 { | ||
| continue | ||
| } | ||
| if ent.Index > state.Commit { | ||
| break | ||
| } | ||
| request, err := parseEntryNormal(ent) | ||
| if err != nil { | ||
| return nil, err | ||
| } | ||
| if request != nil { | ||
| requests = append(requests, *request) | ||
| } | ||
| } | ||
| return requests, nil | ||
| } | ||
|
|
||
| func ReadWAL(lg *zap.Logger, dataDir string) (state raftpb.HardState, ents []raftpb.Entry, err error) { | ||
| walDir := filepath.Join(dataDir, "member", "wal") | ||
| repaired := false | ||
| for { | ||
| w, err := wal.OpenForRead(lg, walDir, walpb.Snapshot{Index: 0}) | ||
| if err != nil { | ||
| lg.Fatal("failed to open WAL", zap.Error(err)) | ||
| } | ||
| _, state, ents, err = w.ReadAll() | ||
| w.Close() | ||
| if err != nil { | ||
| if errors.Is(err, wal.ErrSnapshotNotFound) { | ||
| return state, ents, nil | ||
| } | ||
| // we can only repair ErrUnexpectedEOF and we never repair twice. | ||
| if repaired || !errors.Is(err, io.ErrUnexpectedEOF) { | ||
| return state, nil, fmt.Errorf("failed to read WAL, cannot be repaired, err: %s", err) | ||
| } | ||
| if !wal.Repair(lg, walDir) { | ||
| return state, nil, fmt.Errorf("failed to repair WAL, err: %s", err) | ||
| } else { | ||
| lg.Info("repaired WAL", zap.Error(err)) | ||
| repaired = true | ||
| } | ||
| continue | ||
| } | ||
| return state, ents, nil | ||
| } | ||
| } | ||
|
|
||
| func parseEntryNormal(ent raftpb.Entry) (*model.EtcdRequest, error) { | ||
| var raftReq pb.InternalRaftRequest | ||
| if err := raftReq.Unmarshal(ent.Data); err != nil { | ||
| return nil, err | ||
| } | ||
| switch { | ||
| case raftReq.Put != nil: | ||
| op := model.PutOptions{Key: string(raftReq.Put.Key), Value: model.ToValueOrHash(string(raftReq.Put.Value))} | ||
| request := model.EtcdRequest{ | ||
| Type: model.Txn, | ||
| Txn: &model.TxnRequest{ | ||
| OperationsOnSuccess: []model.EtcdOperation{ | ||
| {Type: model.PutOperation, Put: op}, | ||
| }, | ||
| }, | ||
| } | ||
| return &request, nil | ||
| case raftReq.ClusterMemberAttrSet != nil: | ||
| return nil, nil | ||
| case raftReq.ClusterVersionSet != nil: | ||
| return nil, nil | ||
| case raftReq.Compaction != nil: | ||
| return nil, nil | ||
| case raftReq.Txn != nil: | ||
| txn := model.TxnRequest{} | ||
| if raftReq.Txn.Failure != nil { | ||
| panic("not empty failure operation") | ||
| } | ||
| if raftReq.Txn.Compare != nil { | ||
| panic("not empty conditions") | ||
| } | ||
| for _, op := range raftReq.Txn.Success { | ||
| var operation model.EtcdOperation | ||
| switch { | ||
| case op.GetRequestRange() != nil: | ||
| rangeOp := op.GetRequestRange() | ||
| operation = model.EtcdOperation{ | ||
| Type: model.RangeOperation, | ||
| Range: model.RangeOptions{ | ||
| Start: string(rangeOp.Key), | ||
| End: string(rangeOp.RangeEnd), | ||
| Limit: rangeOp.Limit, | ||
| }, | ||
| } | ||
| case op.GetRequestPut() != nil: | ||
| putOp := op.GetRequestPut() | ||
| operation = model.EtcdOperation{ | ||
| Type: model.PutOperation, | ||
| Put: model.PutOptions{ | ||
| Key: string(putOp.Key), | ||
| Value: model.ToValueOrHash(string(putOp.Value)), | ||
| }, | ||
| } | ||
| case op.GetRequestDeleteRange() != nil: | ||
| deleteOp := op.GetRequestDeleteRange() | ||
| operation = model.EtcdOperation{ | ||
| Type: model.DeleteOperation, | ||
| Delete: model.DeleteOptions{ | ||
| Key: string(deleteOp.Key), | ||
| }, | ||
| } | ||
| default: | ||
| panic(fmt.Sprintf("Unknown op type %v", op)) | ||
| } | ||
| txn.OperationsOnSuccess = append(txn.OperationsOnSuccess, operation) | ||
| } | ||
| request := model.EtcdRequest{ | ||
| Type: model.Txn, | ||
| Txn: &txn, | ||
| } | ||
| return &request, nil | ||
| default: | ||
| panic(fmt.Sprintf("Unhandled raft request: %+v", raftReq)) | ||
| } | ||
| } |
Binary file added
BIN
+2.52 MB
...nessEtcdHighTrafficClusterOfSize1-test-0/member/wal/0000000000000000-0000000000000000.wal
Binary file not shown.
Oops, something went wrong.