Fixes to metricbeat's KVM module (#7793) (#8706)

Improper error handling caused a panic when connection to libvirtd
couldn't be stablished.

Cleaned up error handling a little bit.

Fixes #7792

The default settings for the kvm module didn't work.

Updated to connect to the unix socket by default, and provide a hint on
how to setup access to libvirtd running on remote hosts.

Co-authored-by: Adrian Serrano <[email protected]>

(cherry picked from commit 76d3949)

CHANGELOG.asciidoc

@@ -81,6 +81,7 @@ https://github.com/elastic/beats/compare/v6.4.0...6.x[Check the HEAD diff]
 - Fix issue that would prevent kafka module to find a proper broker when port is not set {pull}8613[8613]
 - Fix range colors in multiple visualizations. {issue}8633[8633] {pull}8634[8634]
 - Fix incorrect header parsing on http metricbeat module {issue}8564[8564] {pull}8585[8585]
+- Fixed a panic when the kvm module cannot establish a connection to libvirtd. {issue}7792[7792].
 
 *Packetbeat*
 

metricbeat/docs/modules/kvm.asciidoc

@@ -23,7 +23,10 @@ metricbeat.modules:
   metricsets: ["dommemstat"]
   enabled: true
   period: 10s
-  hosts: ["localhost"]
+  hosts: ["unix:///var/run/libvirt/libvirt-sock"]
+  # For remote hosts, setup network access in libvirtd.conf
+  # and use the tcp scheme:
+  # hosts: [ "tcp://<host>:16509" ]
 
   # Timeout to connect to Libvirt server
   #timeout: 1s

metricbeat/metricbeat.reference.yml

@@ -426,7 +426,10 @@ metricbeat.modules:
   metricsets: ["dommemstat"]
   enabled: true
   period: 10s
-  hosts: ["localhost"]
+  hosts: ["unix:///var/run/libvirt/libvirt-sock"]
+  # For remote hosts, setup network access in libvirtd.conf
+  # and use the tcp scheme:
+  # hosts: [ "tcp://<host>:16509" ]
 
   # Timeout to connect to Libvirt server
   #timeout: 1s

metricbeat/module/kvm/_meta/config.reference.yml

@@ -2,7 +2,10 @@
   metricsets: ["dommemstat"]
   enabled: true
   period: 10s
-  hosts: ["localhost"]
+  hosts: ["unix:///var/run/libvirt/libvirt-sock"]
+  # For remote hosts, setup network access in libvirtd.conf
+  # and use the tcp scheme:
+  # hosts: [ "tcp://<host>:16509" ]
 
   # Timeout to connect to Libvirt server
   #timeout: 1s

metricbeat/module/kvm/_meta/config.yml

@@ -2,4 +2,4 @@
   #metricsets:
   #  - dommemstat
   period: 10s
-  hosts: ["localhost"]
+  hosts: ["unix:///var/run/libvirt/libvirt-sock"]

metricbeat/module/kvm/dommemstat/dommemstat.go

@@ -18,17 +18,18 @@
 package dommemstat
 
 import (
-	"errors"
 	"net"
 	"net/url"
 	"time"
 
-	"github.com/elastic/beats/libbeat/common"
-	"github.com/elastic/beats/libbeat/common/cfgwarn"
-	"github.com/elastic/beats/metricbeat/mb"
+	"github.com/pkg/errors"
 
 	"github.com/digitalocean/go-libvirt"
 	"github.com/digitalocean/go-libvirt/libvirttest"
+
+	"github.com/elastic/beats/libbeat/common"
+	"github.com/elastic/beats/libbeat/common/cfgwarn"
+	"github.com/elastic/beats/metricbeat/mb"
 )
 
 const (
@@ -101,30 +102,40 @@ func (m *MetricSet) Fetch(report mb.ReporterV2) {
 
 		c, err = net.DialTimeout(u.Scheme, address, m.Timeout)
 		if err != nil {
-			report.Error(err)
+			report.Error(errors.Wrapf(err, "cannot connect to %v", u))
+			return
 		}
 	}
 
 	defer c.Close()
 
 	l := libvirt.New(c)
-	if err := l.Connect(); err != nil {
-		report.Error(err)
+	if err = l.Connect(); err != nil {
+		report.Error(errors.Wrap(err, "error connecting to libvirtd"))
+		return
 	}
+	defer func() {
+		if err = l.Disconnect(); err != nil {
+			report.Error(errors.Wrap(err, "failed to disconnect"))
+		}
+	}()
 
 	domains, err := l.Domains()
 	if err != nil {
-		report.Error(err)
+		report.Error(errors.Wrap(err, "error listing domains"))
+		return
 	}
 
 	for _, d := range domains {
 		gotDomainMemoryStats, err := l.DomainMemoryStats(d, maximumStats, flags)
 		if err != nil {
-			report.Error(err)
+			report.Error(errors.Wrapf(err, "error fetching memory stats for domain %s", d.Name))
+			continue
 		}
 
 		if len(gotDomainMemoryStats) == 0 {
-			report.Error(errors.New("no domain memory stats found"))
+			report.Error(errors.Errorf("no memory stats for domain %s", d.Name))
+			continue
 		}
 
 		for i := range gotDomainMemoryStats {
@@ -140,10 +151,6 @@ func (m *MetricSet) Fetch(report mb.ReporterV2) {
 			})
 		}
 	}
-
-	if err := l.Disconnect(); err != nil {
-		report.Error(errors.New("failed to disconnect"))
-	}
 }
 
 func getDomainMemoryStatName(tag int32) string {

metricbeat/modules.d/kvm.yml.disabled

@@ -5,4 +5,4 @@
   #metricsets:
   #  - dommemstat
   period: 10s
-  hosts: ["localhost"]
+  hosts: ["unix:///var/run/libvirt/libvirt-sock"]