Send health diagnostics data as part of callhome (#16006)

cmd/admin-handlers.go

@@ -1911,67 +1911,7 @@ func getKubernetesInfo(dctx context.Context) madmin.KubernetesInfo {
 	return ki
 }
 
-// HealthInfoHandler - GET /minio/admin/v3/healthinfo
-// ----------
-// Get server health info
-func (a adminAPIHandlers) HealthInfoHandler(w http.ResponseWriter, r *http.Request) {
-	ctx := newContext(r, w, "HealthInfo")
-
-	defer logger.AuditLog(ctx, w, r, mustGetClaimsFromToken(r))
-
-	objectAPI, _ := validateAdminReq(ctx, w, r, iampolicy.HealthInfoAdminAction)
-	if objectAPI == nil {
-		return
-	}
-
-	query := r.Form
-	healthInfo := madmin.HealthInfo{
-		Version: madmin.HealthInfoVersion,
-		Minio: madmin.MinioHealthInfo{
-			Info: madmin.MinioInfo{
-				DeploymentID: globalDeploymentID,
-			},
-		},
-	}
-	healthInfoCh := make(chan madmin.HealthInfo)
-
-	enc := json.NewEncoder(w)
-
-	setCommonHeaders(w)
-
-	setEventStreamHeaders(w)
-
-	w.WriteHeader(http.StatusOK)
-
-	errResp := func(err error) {
-		errorResponse := getAPIErrorResponse(ctx, toAdminAPIErr(ctx, err), r.URL.String(),
-			w.Header().Get(xhttp.AmzRequestID), globalDeploymentID)
-		encodedErrorResponse := encodeResponse(errorResponse)
-		healthInfo.Error = string(encodedErrorResponse)
-		logger.LogIf(ctx, enc.Encode(healthInfo))
-	}
-
-	deadline := 10 * time.Second // Default deadline is 10secs for health diagnostics.
-	if dstr := r.Form.Get("deadline"); dstr != "" {
-		var err error
-		deadline, err = time.ParseDuration(dstr)
-		if err != nil {
-			errResp(err)
-			return
-		}
-	}
-
-	nsLock := objectAPI.NewNSLock(minioMetaBucket, "health-check-in-progress")
-	lkctx, err := nsLock.GetLock(ctx, newDynamicTimeout(deadline, deadline))
-	if err != nil { // returns a locked lock
-		errResp(err)
-		return
-	}
-	defer nsLock.Unlock(lkctx.Cancel)
-
-	healthCtx, healthCancel := context.WithTimeout(lkctx.Context(), deadline)
-	defer healthCancel()
-
+func fetchHealthInfo(healthCtx context.Context, objectAPI ObjectLayer, query *url.Values, healthInfoCh chan madmin.HealthInfo, healthInfo madmin.HealthInfo) {
 	hostAnonymizer := createHostAnonymizer()
 	// anonAddr - Anonymizes hosts in given input string.
 	anonAddr := func(addr string) string {
@@ -2196,7 +2136,7 @@ func (a adminAPIHandlers) HealthInfoHandler(w http.ResponseWriter, r *http.Reque
 
 	getAndWriteMinioConfig := func() {
 		if query.Get("minioconfig") == "true" {
-			config, err := readServerConfig(ctx, objectAPI)
+			config, err := readServerConfig(healthCtx, objectAPI)
 			if err != nil {
 				healthInfo.Minio.Config = madmin.MinioConfig{
 					Error: err.Error(),
@@ -2244,7 +2184,7 @@ func (a adminAPIHandlers) HealthInfoHandler(w http.ResponseWriter, r *http.Reque
 		getAndWriteSysConfig()
 
 		if query.Get("minioinfo") == "true" {
-			infoMessage := getServerInfo(ctx, r)
+			infoMessage := getServerInfo(healthCtx, nil)
 			servers := make([]madmin.ServerInfo, 0, len(infoMessage.Servers))
 			for _, server := range infoMessage.Servers {
 				anonEndpoint := anonAddr(server.Endpoint)
@@ -2294,6 +2234,67 @@ func (a adminAPIHandlers) HealthInfoHandler(w http.ResponseWriter, r *http.Reque
 			partialWrite(healthInfo)
 		}
 	}()
+}
+
+// HealthInfoHandler - GET /minio/admin/v3/healthinfo
+// ----------
+// Get server health info
+func (a adminAPIHandlers) HealthInfoHandler(w http.ResponseWriter, r *http.Request) {
+	ctx := newContext(r, w, "HealthInfo")
+	defer logger.AuditLog(ctx, w, r, mustGetClaimsFromToken(r))
+
+	objectAPI, _ := validateAdminReq(ctx, w, r, iampolicy.HealthInfoAdminAction)
+	if objectAPI == nil {
+		return
+	}
+
+	query := r.Form
+	healthInfoCh := make(chan madmin.HealthInfo)
+	enc := json.NewEncoder(w)
+
+	healthInfo := madmin.HealthInfo{
+		Version: madmin.HealthInfoVersion,
+		Minio: madmin.MinioHealthInfo{
+			Info: madmin.MinioInfo{
+				DeploymentID: globalDeploymentID,
+			},
+		},
+	}
+
+	errResp := func(err error) {
+		errorResponse := getAPIErrorResponse(ctx, toAdminAPIErr(ctx, err), r.URL.String(),
+			w.Header().Get(xhttp.AmzRequestID), globalDeploymentID)
+		encodedErrorResponse := encodeResponse(errorResponse)
+		healthInfo.Error = string(encodedErrorResponse)
+		logger.LogIf(ctx, enc.Encode(healthInfo))
+	}
+
+	deadline := 10 * time.Second // Default deadline is 10secs for health diagnostics.
+	if dstr := query.Get("deadline"); dstr != "" {
+		var err error
+		deadline, err = time.ParseDuration(dstr)
+		if err != nil {
+			errResp(err)
+			return
+		}
+	}
+
+	nsLock := objectAPI.NewNSLock(minioMetaBucket, "health-check-in-progress")
+	lkctx, err := nsLock.GetLock(ctx, newDynamicTimeout(deadline, deadline))
+	if err != nil { // returns a locked lock
+		errResp(err)
+		return
+	}
+
+	defer nsLock.Unlock(lkctx.Cancel)
+	healthCtx, healthCancel := context.WithTimeout(lkctx.Context(), deadline)
+	defer healthCancel()
+
+	go fetchHealthInfo(healthCtx, objectAPI, &query, healthInfoCh, healthInfo)
+
+	setCommonHeaders(w)
+	setEventStreamHeaders(w)
+	w.WriteHeader(http.StatusOK)
 
 	ticker := time.NewTicker(5 * time.Second)
 	defer ticker.Stop()

cmd/callhome.go

@@ -18,29 +18,20 @@
 package cmd
 
 import (
+	"bytes"
+	"compress/gzip"
 	"context"
+	"encoding/json"
+	"errors"
 	"fmt"
 	"math/rand"
+	"net/url"
 	"time"
 
 	"github.com/minio/madmin-go"
 	"github.com/minio/minio/internal/logger"
 )
 
-const (
-	// callhomeSchemaVersion1 is callhome schema version 1
-	callhomeSchemaVersion1 = "1"
-
-	// callhomeSchemaVersion is current callhome schema version.
-	callhomeSchemaVersion = callhomeSchemaVersion1
-)
-
-// CallhomeInfo - Contains callhome information
-type CallhomeInfo struct {
-	SchemaVersion string             `json:"schema_version"`
-	AdminInfo     madmin.InfoMessage `json:"admin_info"`
-}
-
 var callhomeLeaderLockTimeout = newDynamicTimeout(30*time.Second, 10*time.Second)
 
 // initCallhome will start the callhome task in the background.
@@ -118,26 +109,91 @@ func runCallhome(ctx context.Context, objAPI ObjectLayer) bool {
 }
 
 func performCallhome(ctx context.Context) {
-	err := sendCallhomeInfo(
-		CallhomeInfo{
-			SchemaVersion: callhomeSchemaVersion,
-			AdminInfo:     getServerInfo(ctx, nil),
-		})
-	if err != nil {
-		logger.LogIf(ctx, fmt.Errorf("Unable to perform callhome: %w", err))
+	deadline := 10 * time.Second // Default deadline is 10secs for callhome
+	objectAPI := newObjectLayerFn()
+	if objectAPI == nil {
+		logger.LogIf(ctx, errors.New("Callhome: object layer not ready"))
+		return
+	}
+
+	healthCtx, healthCancel := context.WithTimeout(ctx, deadline)
+	defer healthCancel()
+
+	healthInfoCh := make(chan madmin.HealthInfo)
+
+	query := url.Values{}
+	for _, k := range madmin.HealthDataTypesList {
+		query.Set(string(k), "true")
+	}
+
+	healthInfo := madmin.HealthInfo{
+		Version: madmin.HealthInfoVersion,
+		Minio: madmin.MinioHealthInfo{
+			Info: madmin.MinioInfo{
+				DeploymentID: globalDeploymentID,
+			},
+		},
+	}
+
+	go fetchHealthInfo(healthCtx, objectAPI, &query, healthInfoCh, healthInfo)
+
+	for {
+		select {
+		case hi, hasMore := <-healthInfoCh:
+			if !hasMore {
+				// Received all data. Send to SUBNET and return
+				err := sendHealthInfo(ctx, healthInfo)
+				if err != nil {
+					logger.LogIf(ctx, fmt.Errorf("Unable to perform callhome: %w", err))
+				}
+				return
+			}
+			healthInfo = hi
+		case <-healthCtx.Done():
+			return
+		}
 	}
 }
 
 const (
-	callhomeURL    = "https://subnet.min.io/api/callhome"
-	callhomeURLDev = "http://localhost:9000/api/callhome"
+	healthURL    = "https://subnet.min.io/api/health/upload"
+	healthURLDev = "http://localhost:9000/api/health/upload"
 )
 
-func sendCallhomeInfo(ch CallhomeInfo) error {
-	url := callhomeURL
+func sendHealthInfo(ctx context.Context, healthInfo madmin.HealthInfo) error {
+	url := healthURL
 	if globalIsCICD {
-		url = callhomeURLDev
+		url = healthURLDev
 	}
-	_, err := globalSubnetConfig.Post(url, ch)
+
+	filename := fmt.Sprintf("health_%s.json.gz", UTCNow().Format("20060102150405"))
+	url += "?filename=" + filename
+
+	_, err := globalSubnetConfig.Upload(url, filename, createHealthJSONGzip(ctx, healthInfo))
 	return err
 }
+
+func createHealthJSONGzip(ctx context.Context, healthInfo madmin.HealthInfo) []byte {
+	var b bytes.Buffer
+	gzWriter := gzip.NewWriter(&b)
+
+	header := struct {
+		Version string `json:"version"`
+	}{Version: healthInfo.Version}
+
+	enc := json.NewEncoder(gzWriter)
+	if e := enc.Encode(header); e != nil {
+		logger.LogIf(ctx, fmt.Errorf("Could not encode health info header: %w", e))
+		return nil
+	}
+
+	if e := enc.Encode(healthInfo); e != nil {
+		logger.LogIf(ctx, fmt.Errorf("Could not encode health info: %w", e))
+		return nil
+	}
+
+	gzWriter.Flush()
+	gzWriter.Close()
+
+	return b.Bytes()
+}

internal/config/subnet/subnet.go

@@ -23,6 +23,7 @@ import (
 	"errors"
 	"fmt"
 	"io"
+	"mime/multipart"
 	"net/http"
 	"time"
 
@@ -33,25 +34,38 @@ const (
 	respBodyLimit = 1 << 20 // 1 MiB
 )
 
-// Post submit 'payload' to specified URL
-func (c Config) Post(reqURL string, payload interface{}) (string, error) {
+// Upload given file content (payload) to specified URL
+func (c Config) Upload(reqURL string, filename string, payload []byte) (string, error) {
 	if !c.Registered() {
 		return "", errors.New("Deployment is not registered with SUBNET. Please register the deployment via 'mc license register ALIAS'")
 	}
-	body, err := json.Marshal(payload)
-	if err != nil {
-		return "", err
-	}
-	r, err := http.NewRequest(http.MethodPost, reqURL, bytes.NewReader(body))
-	if err != nil {
-		return "", err
+
+	var body bytes.Buffer
+	writer := multipart.NewWriter(&body)
+	part, e := writer.CreateFormFile("file", filename)
+	if e != nil {
+		return "", e
 	}
 
+	if _, e = part.Write(payload); e != nil {
+		return "", e
+	}
+	writer.Close()
+
+	r, e := http.NewRequest(http.MethodPost, reqURL, &body)
+	if e != nil {
+		return "", e
+	}
+	r.Header.Add("Content-Type", writer.FormDataContentType())
+
+	return c.submitPost(r)
+}
+
+func (c Config) submitPost(r *http.Request) (string, error) {
 	configLock.RLock()
-	r.Header.Set("Authorization", c.APIKey)
+	r.Header.Set(xhttp.SubnetAPIKey, c.APIKey)
 	configLock.RUnlock()
-
-	r.Header.Set("Content-Type", "application/json")
+	r.Header.Set(xhttp.MinioDeploymentID, xhttp.GlobalDeploymentID)
 
 	client := &http.Client{
 		Timeout:   10 * time.Second,
@@ -76,3 +90,22 @@ func (c Config) Post(reqURL string, payload interface{}) (string, error) {
 
 	return respStr, fmt.Errorf("SUBNET request failed with code %d and error: %s", resp.StatusCode, respStr)
 }
+
+// Post submit 'payload' to specified URL
+func (c Config) Post(reqURL string, payload interface{}) (string, error) {
+	if !c.Registered() {
+		return "", errors.New("Deployment is not registered with SUBNET. Please register the deployment via 'mc license register ALIAS'")
+	}
+	body, err := json.Marshal(payload)
+	if err != nil {
+		return "", err
+	}
+	r, err := http.NewRequest(http.MethodPost, reqURL, bytes.NewReader(body))
+	if err != nil {
+		return "", err
+	}
+
+	r.Header.Set("Content-Type", "application/json")
+
+	return c.submitPost(r)
+}

internal/http/headers.go

@@ -210,6 +210,9 @@ const (
 
 	// MinIOCompressed is returned when object is compressed
 	MinIOCompressed = "X-Minio-Compressed"
+
+	// SUBNET related
+	SubnetAPIKey = "x-subnet-api-key"
 )
 
 // Common http query params S3 API