improve obd tests and optimize network (#9378)

- keep long running obd network tests alive
- fix error - wrong number of parents in process OBD info
- ensure that osinfo does not error out when inside containers
- remove limit on max number of connections per client transport

The generic client transport uses a default limit of 64 conns per transport.
This could end up limiting and throttling usage, and artificially slowing
down the performance of MinIO even on hardware capable of doing better.

cmd/obdinfo.go

@@ -317,16 +317,14 @@ func getLocalProcOBD(ctx context.Context) madmin.ServerProcOBDInfo {
 		sysProc.PageFaults = pageFaults
 
 		parent, err := proc.ParentWithContext(ctx)
-		if err != nil {
-			return errProcInfo(err)
+		if err == nil {
+			sysProc.Parent = parent.Pid
 		}
-		sysProc.Parent = parent.Pid
 
 		ppid, err := proc.PpidWithContext(ctx)
-		if err != nil {
-			return errProcInfo(err)
+		if err == nil {
+			sysProc.Ppid = ppid
 		}
-		sysProc.Ppid = ppid
 
 		rlimit, err := proc.RlimitWithContext(ctx)
 		if err != nil {
@@ -404,13 +402,8 @@ func getLocalOsInfoOBD(ctx context.Context) madmin.ServerOsOBDInfo {
 		}
 	}
 
-	users, err := host.UsersWithContext(ctx)
-	if err != nil {
-		return madmin.ServerOsOBDInfo{
-			Addr:  addr,
-			Error: err.Error(),
-		}
-	}
+	// ignore user err, as it cannot be obtained reliably inside containers
+	users, _ := host.UsersWithContext(ctx)
 
 	return madmin.ServerOsOBDInfo{
 		Addr:    addr,

cmd/peer-rest-client.go

@@ -374,7 +374,11 @@ func (client *peerRESTClient) DispatchNetOBDInfo(ctx context.Context) (info madm
 		return
 	}
 	defer http.DrainBody(respBody)
-	err = gob.NewDecoder(respBody).Decode(&info)
+	waitReader, err := waitForHTTPResponse(respBody)
+	if err != nil {
+		return
+	}
+	err = gob.NewDecoder(waitReader).Decode(&info)
 	return
 }
 

cmd/peer-rest-server.go

@@ -403,9 +403,12 @@ func (s *peerRESTServer) DispatchNetOBDInfoHandler(w http.ResponseWriter, r *htt
 		return
 	}
 
+	done := keepHTTPResponseAlive(w)
+
 	ctx := newContext(r, w, "DispatchNetOBDInfo")
 	info := globalNotificationSys.NetOBDInfo(ctx)
 
+	done()
 	logger.LogIf(ctx, gob.NewEncoder(w).Encode(info))
 	w.(http.Flusher).Flush()
 }

cmd/utils.go

@@ -471,7 +471,6 @@ func newCustomHTTPTransport(tlsConfig *tls.Config, dialTimeout time.Duration) fu
 		DialContext:           newCustomDialContext(dialTimeout, 15*time.Second),
 		MaxIdleConnsPerHost:   16,
 		MaxIdleConns:          16,
-		MaxConnsPerHost:       64, // This is used per drive/rpc host. More requests will block until free.
 		IdleConnTimeout:       1 * time.Minute,
 		ResponseHeaderTimeout: 3 * time.Minute, // Set conservative timeouts for MinIO internode.
 		TLSHandshakeTimeout:   10 * time.Second,