Skip collectors that fail at init. Only write metrics with fields

metric-collector.go

@@ -2,6 +2,7 @@ package main
 
 import (
 	"encoding/json"
+	"flag"
 	"fmt"
 	"github.com/ClusterCockpit/cc-metric-collector/collectors"
 	"github.com/ClusterCockpit/cc-metric-collector/sinks"
@@ -11,7 +12,6 @@ import (
 	"strings"
 	"sync"
 	"time"
-	"flag"
 )
 
 // List of provided collectors. Which collector should be run can be
@@ -25,6 +25,7 @@ var Collectors = map[string]collectors.MetricGetter{
 	"lustrestat": &collectors.LustreCollector{},
 	"cpustat":    &collectors.CpustatCollector{},
 	"topprocs":   &collectors.TopProcsCollector{},
+	"nvidia":     &collectors.NvidiaCollector{},
 }
 
 var Sinks = map[string]sinks.SinkFuncs{
@@ -75,7 +76,7 @@ func ReadCli() map[string]string {
 func SetLogging(logfile string) error {
 	var file *os.File
 	var err error
-    if (logfile != "stderr") {
+	if logfile != "stderr" {
 		file, err = os.OpenFile(logfile, os.O_APPEND|os.O_CREATE|os.O_WRONLY, 0600)
 		if err != nil {
 			log.Fatal(err)
@@ -119,14 +120,14 @@ func main() {
 	}
 	clicfg := ReadCli()
 	err = SetLogging(clicfg["logfile"])
-	if (err != nil) {
+	if err != nil {
 		log.Print("Error setting up logging system to ", clicfg["logfile"])
 		return
 	}
 
 	// Load and check configuration
 	err = LoadConfiguration(clicfg["configfile"], &config)
-	if (err != nil) {
+	if err != nil {
 		log.Print("Error reading configuration file ", clicfg["configfile"])
 		return
 	}
@@ -167,11 +168,18 @@ func main() {
 	shutdown(&wg, &config, sink)
 
 	// Initialize all collectors
+	tmp := make([]string, 0)
 	for _, c := range config.Collectors {
 		col := Collectors[c]
-		col.Init()
+		err = col.Init()
+		if err != nil {
+			log.Print("SKIP ", col.Name())
+		} else {
 			log.Print("Start ", col.Name())
+			tmp = append(tmp, c)
 		}
+	}
+	config.Collectors = tmp
 
 	// Setup up ticker loop
 	log.Print("Running loop every ", time.Duration(config.Interval)*time.Second)
@@ -229,23 +237,29 @@ func main() {
 				}
 
 				// Send out node metrics
+				if len(nodeFields) > 0 {
 					sink.Write("node", map[string]string{"host": host}, nodeFields, t)
+				}
 
 				// Send out socket metrics (if any)
 				if scount > 0 {
 					for sid, socket := range socketsFields {
+						if len(socket) > 0 {
 							sink.Write("socket", map[string]string{"socket": fmt.Sprintf("%d", sid), "host": host}, socket, t)
 						}
 					}
+				}
 
 				// Send out CPU metrics (if any)
 				if ccount > 0 {
 					for cid, cpu := range cpuFields {
+						if len(cpu) > 0 {
 							sink.Write("cpu", map[string]string{"cpu": fmt.Sprintf("%d", cid), "host": host}, cpu, t)
 						}
 					}
 				}
 			}
+		}
 	}()
 
 	// Wait until receiving an interrupt