chore: health endpoint related changes (#2275)

deploy/docker-swarm/clickhouse-setup/docker-compose.yaml

@@ -156,7 +156,7 @@ services:
       - TELEMETRY_ENABLED=true
       - DEPLOYMENT_TYPE=docker-swarm
     healthcheck:
-      test: ["CMD", "wget", "--spider", "-q", "localhost:8080/api/v1/version"]
+      test: ["CMD", "wget", "--spider", "-q", "localhost:8080/api/v1/health"]
       interval: 30s
       timeout: 5s
       retries: 3

deploy/docker/clickhouse-setup/docker-compose-local.yaml

@@ -28,7 +28,7 @@ services:
       - "8080:8080"
     restart: on-failure
     healthcheck:
-      test: ["CMD", "wget", "--spider", "-q", "localhost:8080/api/v1/version"]
+      test: ["CMD", "wget", "--spider", "-q", "localhost:8080/api/v1/health"]
       interval: 30s
       timeout: 5s
       retries: 3

deploy/docker/clickhouse-setup/docker-compose.yaml

@@ -174,7 +174,7 @@ services:
       - DEPLOYMENT_TYPE=docker-standalone-amd
     restart: on-failure
     healthcheck:
-      test: ["CMD", "wget", "--spider", "-q", "localhost:8080/api/v1/version"]
+      test: ["CMD", "wget", "--spider", "-q", "localhost:8080/api/v1/health"]
       interval: 30s
       timeout: 5s
       retries: 3

pkg/query-service/app/http_handler.go

@@ -1704,7 +1704,7 @@ func (aH *APIHandler) getHealth(w http.ResponseWriter, r *http.Request) {
 	if ok {
 		err := aH.reader.CheckClickHouse(r.Context())
 		if err != nil {
-			aH.HandleError(w, err, http.StatusServiceUnavailable)
+			RespondError(w, &model.ApiError{Err: err, Typ: model.ErrorStatusServiceUnavailable}, nil)
 			return
 		}
 	}

pkg/query-service/model/response.go

@@ -49,19 +49,20 @@ func (a *ApiError) IsNil() bool {
 type ErrorType string
 
 const (
-	ErrorNone                  ErrorType = ""
+	ErrorNone                     ErrorType = ""
-	ErrorTimeout               ErrorType = "timeout"
+	ErrorTimeout                  ErrorType = "timeout"
-	ErrorCanceled              ErrorType = "canceled"
+	ErrorCanceled                 ErrorType = "canceled"
-	ErrorExec                  ErrorType = "execution"
+	ErrorExec                     ErrorType = "execution"
-	ErrorBadData               ErrorType = "bad_data"
+	ErrorBadData                  ErrorType = "bad_data"
-	ErrorInternal              ErrorType = "internal"
+	ErrorInternal                 ErrorType = "internal"
-	ErrorUnavailable           ErrorType = "unavailable"
+	ErrorUnavailable              ErrorType = "unavailable"
-	ErrorNotFound              ErrorType = "not_found"
+	ErrorNotFound                 ErrorType = "not_found"
-	ErrorNotImplemented        ErrorType = "not_implemented"
+	ErrorNotImplemented           ErrorType = "not_implemented"
-	ErrorUnauthorized          ErrorType = "unauthorized"
+	ErrorUnauthorized             ErrorType = "unauthorized"
-	ErrorForbidden             ErrorType = "forbidden"
+	ErrorForbidden                ErrorType = "forbidden"
-	ErrorConflict              ErrorType = "conflict"
+	ErrorConflict                 ErrorType = "conflict"
-	ErrorStreamingNotSupported ErrorType = "streaming is not supported"
+	ErrorStreamingNotSupported    ErrorType = "streaming is not supported"
+	ErrorStatusServiceUnavailable ErrorType = "service unavailable"
 )
 
 // BadRequest returns a ApiError object of bad request

pkg/query-service/telemetry/ignored.go

@@ -6,7 +6,7 @@ func IgnoredPaths() map[string]struct{} {
 		"/api/v1/version":             {},
 		"/api/v1/query_range":         {},
 		"/api/v2/metrics/query_range": {},
-		"/api/v1/services/list":       {},
+		"/api/v1/health":              {},
 	}
 
 	return ignoredPaths

pkg/query-service/tests/docker.go

@@ -90,7 +90,7 @@ func startCluster() error {
 
 	client := http.Client{}
 	for i := 0; i < 10; i++ {
-		if _, err := client.Get("http://localhost:8180/api/v1/version"); err != nil {
+		if _, err := client.Get("http://localhost:8180/api/v1/health"); err != nil {
 			time.Sleep(2 * time.Second)
 		} else {
 			log.Printf("CLUSTER UP\n")

pkg/query-service/tests/test-deploy/docker-compose.yaml

@@ -162,7 +162,7 @@ services:
       - DEPLOYMENT_TYPE=docker-standalone-amd
     restart: on-failure
     healthcheck:
-      test: ["CMD", "wget", "--spider", "-q", "localhost:8080/api/v1/version"]
+      test: ["CMD", "wget", "--spider", "-q", "localhost:8080/api/v1/health"]
       interval: 30s
       timeout: 5s
       retries: 3