ClusterCockpit/cc-metric-collector
1
0
Fork 0
mirror of https://github.com/ClusterCockpit/cc-metric-collector.git synced 2024-11-10 04:27:25 +01:00
Code Issues Packages Projects Releases Wiki Activity

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

Browse Source
This commit is contained in:
Thomas Roehl 2021-05-14 19:22:42 +02:00
parent 6b6e28d4c7
commit 1da906470d
1 changed files with 51 additions and 37 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

24
metric-collector.go
View File

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