X-Git-Url: https://git.arvados.org/arvados.git/blobdiff_plain/ce4285dd9a6310a799b861237918273329390316..72d7d41944006d1f48f570784dafe56b9812b0c8:/sdk/go/keepclient/support.go diff --git a/sdk/go/keepclient/support.go b/sdk/go/keepclient/support.go index e589593fa8..8d299815b2 100644 --- a/sdk/go/keepclient/support.go +++ b/sdk/go/keepclient/support.go @@ -5,6 +5,8 @@ package keepclient import ( + "bytes" + "context" "crypto/md5" "errors" "fmt" @@ -13,12 +15,15 @@ import ( "log" "net/http" "os" + "strconv" "strings" - "git.curoverse.com/arvados.git/sdk/go/arvadosclient" + "git.arvados.org/arvados.git/sdk/go/arvados" + "git.arvados.org/arvados.git/sdk/go/arvadosclient" + "git.arvados.org/arvados.git/sdk/go/asyncbuf" ) -// Function used to emit debug messages. The easiest way to enable +// DebugPrintf emits debug messages. The easiest way to enable // keepclient debug messages in your application is to assign // log.Printf to DebugPrintf. var DebugPrintf = func(string, ...interface{}) {} @@ -48,26 +53,27 @@ type svcList struct { } type uploadStatus struct { - err error - url string - statusCode int - replicas_stored int - response string + err error + url string + statusCode int + replicasStored int + classesStored map[string]int + response string } -func (this *KeepClient) uploadToKeepServer(host string, hash string, body io.Reader, - upload_status chan<- uploadStatus, expectedLength int64, reqid string) { +func (kc *KeepClient) uploadToKeepServer(host string, hash string, classesTodo []string, body io.Reader, + uploadStatusChan chan<- uploadStatus, expectedLength int, reqid string) { var req *http.Request var err error var url = fmt.Sprintf("%s/%s", host, hash) if req, err = http.NewRequest("PUT", url, nil); err != nil { DebugPrintf("DEBUG: [%s] Error creating request PUT %v error: %v", reqid, url, err.Error()) - upload_status <- uploadStatus{err, url, 0, 0, ""} + uploadStatusChan <- uploadStatus{err, url, 0, 0, nil, ""} return } - req.ContentLength = expectedLength + req.ContentLength = int64(expectedLength) if expectedLength > 0 { req.Body = ioutil.NopCloser(body) } else { @@ -77,24 +83,29 @@ func (this *KeepClient) uploadToKeepServer(host string, hash string, body io.Rea } req.Header.Add("X-Request-Id", reqid) - req.Header.Add("Authorization", "OAuth2 "+this.Arvados.ApiToken) + req.Header.Add("Authorization", "OAuth2 "+kc.Arvados.ApiToken) req.Header.Add("Content-Type", "application/octet-stream") - req.Header.Add(X_Keep_Desired_Replicas, fmt.Sprint(this.Want_replicas)) - if len(this.StorageClasses) > 0 { - req.Header.Add("X-Keep-Storage-Classes", strings.Join(this.StorageClasses, ", ")) + req.Header.Add(XKeepDesiredReplicas, fmt.Sprint(kc.Want_replicas)) + if len(classesTodo) > 0 { + req.Header.Add(XKeepStorageClasses, strings.Join(classesTodo, ", ")) } var resp *http.Response - if resp, err = this.httpClient().Do(req); err != nil { + if resp, err = kc.httpClient().Do(req); err != nil { DebugPrintf("DEBUG: [%s] Upload failed %v error: %v", reqid, url, err.Error()) - upload_status <- uploadStatus{err, url, 0, 0, err.Error()} + uploadStatusChan <- uploadStatus{err, url, 0, 0, nil, err.Error()} return } rep := 1 - if xr := resp.Header.Get(X_Keep_Replicas_Stored); xr != "" { + if xr := resp.Header.Get(XKeepReplicasStored); xr != "" { fmt.Sscanf(xr, "%d", &rep) } + scc := resp.Header.Get(XKeepStorageClassesConfirmed) + classesStored, err := parseStorageClassesConfirmedHeader(scc) + if err != nil { + DebugPrintf("DEBUG: [%s] Ignoring invalid %s header %q: %s", reqid, XKeepStorageClassesConfirmed, scc, err) + } defer resp.Body.Close() defer io.Copy(ioutil.Discard, resp.Body) @@ -103,75 +114,144 @@ func (this *KeepClient) uploadToKeepServer(host string, hash string, body io.Rea response := strings.TrimSpace(string(respbody)) if err2 != nil && err2 != io.EOF { DebugPrintf("DEBUG: [%s] Upload %v error: %v response: %v", reqid, url, err2.Error(), response) - upload_status <- uploadStatus{err2, url, resp.StatusCode, rep, response} + uploadStatusChan <- uploadStatus{err2, url, resp.StatusCode, rep, classesStored, response} } else if resp.StatusCode == http.StatusOK { DebugPrintf("DEBUG: [%s] Upload %v success", reqid, url) - upload_status <- uploadStatus{nil, url, resp.StatusCode, rep, response} + uploadStatusChan <- uploadStatus{nil, url, resp.StatusCode, rep, classesStored, response} } else { if resp.StatusCode >= 300 && response == "" { response = resp.Status } DebugPrintf("DEBUG: [%s] Upload %v error: %v response: %v", reqid, url, resp.StatusCode, response) - upload_status <- uploadStatus{errors.New(resp.Status), url, resp.StatusCode, rep, response} + uploadStatusChan <- uploadStatus{errors.New(resp.Status), url, resp.StatusCode, rep, classesStored, response} } } -func (this *KeepClient) putReplicas( - hash string, - getReader func() io.Reader, - expectedLength int64) (locator string, replicas int, err error) { - - reqid := this.getRequestID() +func (kc *KeepClient) BlockWrite(ctx context.Context, req arvados.BlockWriteOptions) (arvados.BlockWriteResponse, error) { + var resp arvados.BlockWriteResponse + var getReader func() io.Reader + if req.Data == nil && req.Reader == nil { + return resp, errors.New("invalid BlockWriteOptions: Data and Reader are both nil") + } + if req.DataSize < 0 { + return resp, fmt.Errorf("invalid BlockWriteOptions: negative DataSize %d", req.DataSize) + } + if req.DataSize > BLOCKSIZE || len(req.Data) > BLOCKSIZE { + return resp, ErrOversizeBlock + } + if req.Data != nil { + if req.DataSize > len(req.Data) { + return resp, errors.New("invalid BlockWriteOptions: DataSize > len(Data)") + } + if req.DataSize == 0 { + req.DataSize = len(req.Data) + } + getReader = func() io.Reader { return bytes.NewReader(req.Data[:req.DataSize]) } + } else { + buf := asyncbuf.NewBuffer(make([]byte, 0, req.DataSize)) + go func() { + _, err := io.Copy(buf, HashCheckingReader{req.Reader, md5.New(), req.Hash}) + buf.CloseWithError(err) + }() + getReader = buf.NewReader + } + if req.Hash == "" { + m := md5.New() + _, err := io.Copy(m, getReader()) + if err != nil { + return resp, err + } + req.Hash = fmt.Sprintf("%x", m.Sum(nil)) + } + if req.StorageClasses == nil { + if len(kc.StorageClasses) > 0 { + req.StorageClasses = kc.StorageClasses + } else { + req.StorageClasses = kc.DefaultStorageClasses + } + } + if req.Replicas == 0 { + req.Replicas = kc.Want_replicas + } + if req.RequestID == "" { + req.RequestID = kc.getRequestID() + } + if req.Attempts == 0 { + req.Attempts = 1 + kc.Retries + } // Calculate the ordering for uploading to servers - sv := NewRootSorter(this.WritableLocalRoots(), hash).GetSortedRoots() + sv := NewRootSorter(kc.WritableLocalRoots(), req.Hash).GetSortedRoots() // The next server to try contacting - next_server := 0 + nextServer := 0 // The number of active writers active := 0 // Used to communicate status from the upload goroutines - upload_status := make(chan uploadStatus) + uploadStatusChan := make(chan uploadStatus) defer func() { // Wait for any abandoned uploads (e.g., we started // two uploads and the first replied with replicas=2) // to finish before closing the status channel. go func() { for active > 0 { - <-upload_status + <-uploadStatusChan } - close(upload_status) + close(uploadStatusChan) }() }() - replicasDone := 0 - replicasTodo := this.Want_replicas + replicasTodo := map[string]int{} + for _, c := range req.StorageClasses { + replicasTodo[c] = req.Replicas + } - replicasPerThread := this.replicasPerService + replicasPerThread := kc.replicasPerService if replicasPerThread < 1 { // unlimited or unknown - replicasPerThread = replicasTodo + replicasPerThread = req.Replicas } - retriesRemaining := 1 + this.Retries + retriesRemaining := req.Attempts var retryServers []string lastError := make(map[string]string) + trackingClasses := len(replicasTodo) > 0 for retriesRemaining > 0 { - retriesRemaining -= 1 - next_server = 0 + retriesRemaining-- + nextServer = 0 retryServers = []string{} - for replicasTodo > 0 { - for active*replicasPerThread < replicasTodo { + for { + var classesTodo []string + var maxConcurrency int + for sc, r := range replicasTodo { + classesTodo = append(classesTodo, sc) + if maxConcurrency == 0 || maxConcurrency > r { + // Having more than r + // writes in flight + // would overreplicate + // class sc. + maxConcurrency = r + } + } + if !trackingClasses { + maxConcurrency = req.Replicas - resp.Replicas + } + if maxConcurrency < 1 { + // If there are no non-zero entries in + // replicasTodo, we're done. + break + } + for active*replicasPerThread < maxConcurrency { // Start some upload requests - if next_server < len(sv) { - DebugPrintf("DEBUG: [%s] Begin upload %s to %s", reqid, hash, sv[next_server]) - go this.uploadToKeepServer(sv[next_server], hash, getReader(), upload_status, expectedLength, reqid) - next_server += 1 - active += 1 + if nextServer < len(sv) { + DebugPrintf("DEBUG: [%s] Begin upload %s to %s", req.RequestID, req.Hash, sv[nextServer]) + go kc.uploadToKeepServer(sv[nextServer], req.Hash, classesTodo, getReader(), uploadStatusChan, req.DataSize, req.RequestID) + nextServer++ + active++ } else { if active == 0 && retriesRemaining == 0 { msg := "Could not write sufficient replicas: " @@ -179,47 +259,85 @@ func (this *KeepClient) putReplicas( msg += resp + "; " } msg = msg[:len(msg)-2] - return locator, replicasDone, InsufficientReplicasError(errors.New(msg)) - } else { - break + return resp, InsufficientReplicasError{error: errors.New(msg)} } + break } } - DebugPrintf("DEBUG: [%s] Replicas remaining to write: %v active uploads: %v", - reqid, replicasTodo, active) - - // Now wait for something to happen. - if active > 0 { - status := <-upload_status - active -= 1 - - if status.statusCode == 200 { - // good news! - replicasDone += status.replicas_stored - replicasTodo -= status.replicas_stored - locator = status.response - delete(lastError, status.url) - } else { - msg := fmt.Sprintf("[%d] %s", status.statusCode, status.response) - if len(msg) > 100 { - msg = msg[:100] - } - lastError[status.url] = msg - } - if status.statusCode == 0 || status.statusCode == 408 || status.statusCode == 429 || - (status.statusCode >= 500 && status.statusCode != 503) { - // Timeout, too many requests, or other server side failure - // Do not retry when status code is 503, which means the keep server is full - retryServers = append(retryServers, status.url[0:strings.LastIndex(status.url, "/")]) + DebugPrintf("DEBUG: [%s] Replicas remaining to write: %v active uploads: %v", req.RequestID, replicasTodo, active) + if active < 1 { + break + } + + // Wait for something to happen. + status := <-uploadStatusChan + active-- + + if status.statusCode == http.StatusOK { + delete(lastError, status.url) + resp.Replicas += status.replicasStored + if len(status.classesStored) == 0 { + // Server doesn't report + // storage classes. Give up + // trying to track which ones + // are satisfied; just rely on + // total # replicas. + trackingClasses = false + } + for className, replicas := range status.classesStored { + if replicasTodo[className] > replicas { + replicasTodo[className] -= replicas + } else { + delete(replicasTodo, className) + } } + resp.Locator = status.response } else { - break + msg := fmt.Sprintf("[%d] %s", status.statusCode, status.response) + if len(msg) > 100 { + msg = msg[:100] + } + lastError[status.url] = msg + } + + if status.statusCode == 0 || status.statusCode == 408 || status.statusCode == 429 || + (status.statusCode >= 500 && status.statusCode != 503) { + // Timeout, too many requests, or other server side failure + // Do not retry when status code is 503, which means the keep server is full + retryServers = append(retryServers, status.url[0:strings.LastIndex(status.url, "/")]) } } sv = retryServers } - return locator, replicasDone, nil + return resp, nil +} + +func parseStorageClassesConfirmedHeader(hdr string) (map[string]int, error) { + if hdr == "" { + return nil, nil + } + classesStored := map[string]int{} + for _, cr := range strings.Split(hdr, ",") { + cr = strings.TrimSpace(cr) + if cr == "" { + continue + } + fields := strings.SplitN(cr, "=", 2) + if len(fields) != 2 { + return nil, fmt.Errorf("expected exactly one '=' char in entry %q", cr) + } + className := fields[0] + if className == "" { + return nil, fmt.Errorf("empty class name in entry %q", cr) + } + replicas, err := strconv.Atoi(fields[1]) + if err != nil || replicas < 1 { + return nil, fmt.Errorf("invalid replica count %q", fields[1]) + } + classesStored[className] = replicas + } + return classesStored, nil }