feat: new collector for number of processes/threads per user

METRICS.md

@@ -40,3 +40,10 @@
 	- io_stalled_avg60 (float)
 	- memory_waiting_avg60 (float) percentage over 60s
 	- memory_stalled_avg60 (float) percentage over 60s
+- nodestat_userprocs
+  - tags:
+    - group
+	- user
+  - fields:
+    - processes (int)
+	- threads (int)

README.md

@@ -36,6 +36,7 @@ Current collectors are:\
  fc_host - fibrechannel metrics from /sys/class/fc_host/\
  net - network interface metrics from /sys/class/net/\
  pressure - metrics from /proc/pressure/\
+ userprocs - metrics for the number of processes and threads per user\
 Metric timestamp precision will be 1s.
 
 * Restart or reload Telegraf.
@@ -54,6 +55,8 @@ nodestat_fc_host,nodename=20000025ff1bab79,type=NPort\ (fabric\ via\ point-to-po
 nodestat_net,interface=eno1,protocol=ethernet carrier=1i,flag_running=true,flag_up=true,operstate_code=0i,flag_lower_up=true,dormant=0i,duplex="full",link_mode=0i,operstate="up"  1662965695000000000
 nodestat_net,interface=enp3s0f0,protocol=ethernet flag_running=true,flag_up=true,carrier=0i,duplex="unknown",link_mode=0i,operstate="down",operstate_code=5i,flag_lower_up=true,dormant=0i 1662965695000000000
 nodestat_pressure cpu_waiting_avg60=0,io_waiting_avg60=21.1,io_stalled_avg60=20.97 1662965695000000000
+nodestat_userprocs,group=root,user=root threads=173i,processes=124i 1662965695000000000
+nodestat_userprocs,group=postfix,user=postfix processes=3i,threads=3i 1662965695000000000
 ```
 
 # Metrics

internal/collectors/collectors.go

@@ -25,7 +25,7 @@ type CollectorInfo struct {
 var collectInfos []CollectorInfo
 
 func init() {
-	var c []CollectorInfo
+	var c = make([]CollectorInfo, 0, 4)
 	var ci CollectorInfo
 
 	ci = CollectorInfo{"fc_host", true, "fibrechannels", fssys.GatherSysFcHostInfo}
@@ -34,6 +34,8 @@ func init() {
 	c = append(c, ci)
 	ci = CollectorInfo{"pressure", true, "pressure", fsproc.GatherProcPressureInfo}
 	c = append(c, ci)
+	ci = CollectorInfo{"userprocs", true, "processes per user", fsproc.GatherProcUserProcsInfo}
+	c = append(c, ci)
 	collectInfos = c
 }
 

internal/collectors/fsproc/userprocs.go

@@ -0,0 +1,91 @@
+// fsproc functions show metrics from linux /proc filesystem using influx line protocol
+//
+// Author: Tesifonte Belda
+// License: The MIT License (MIT)
+//
+// References:
+//  https://github.com/prometheus/node_exporter/tree/master/collector/pressure_linux.go
+
+package fsproc
+
+import (
+	"fmt"
+	"os"
+	"os/user"
+	"time"
+
+	"github.com/prometheus/procfs"
+	"github.com/tesibelda/lightmetric/metric"
+)
+
+type userInfo struct {
+	processes int
+	threads   int
+}
+
+// GatherProcUserProcsInfo prints number of process per user metrics from /proc/<PID>/status>
+func GatherProcUserProcsInfo() error {
+	fs, err := procfs.NewDefaultFS()
+	if err != nil {
+		return fmt.Errorf("failed to open procfs: %w", err)
+	}
+
+	p, err := fs.AllProcs()
+	if err != nil {
+		return fmt.Errorf("unable to list all processes: %w", err)
+	}
+
+	uidprocs := make(map[string]userInfo, 10)
+	totalProcs, th, ok := 0, 0, false
+	info := userInfo{}
+	for _, pid := range p {
+		status, err := pid.NewStatus()
+		if err != nil {
+			// PIDs can vanish between getting the list and getting stats.
+			continue
+		}
+
+		th = 0
+		stat, err := pid.Stat()
+		if err == nil {
+			th = stat.NumThreads
+		}
+
+		if info, ok = uidprocs[status.UIDs[0]]; !ok {
+			info = userInfo{}
+		}
+		info.processes++
+		info.threads += th
+		uidprocs[status.UIDs[0]] = info
+		totalProcs++
+	}
+	if totalProcs == 0 {
+		return fmt.Errorf("unable to list any processes")
+	}
+
+	fields := make(map[string]interface{}, 2)
+	tags := make(map[string]string, 2)
+	var t time.Time
+	var m metric.Metric
+	for k, v := range uidprocs {
+		usr, err := user.LookupId(k)
+		if err != nil {
+			continue
+		}
+		if len(usr.Username) > 0 {
+			grp, err := user.LookupGroupId(usr.Gid)
+			if err != nil {
+				grp = &user.Group{}
+			}
+
+			fields["processes"] = v.processes
+			fields["threads"] = v.threads
+			tags["user"] = usr.Username
+			tags["group"] = grp.Name
+			t = metric.TimeWithPrecision(time.Now(), time.Second)
+			m = metric.New("nodestat_userprocs", tags, fields, t)
+			fmt.Fprint(os.Stdout, m.String(metric.InfluxLp))
+		}
+	}
+	return nil
+}