metrics: Add virtiofsd exporter

Export proc stats for virtiofsd.

This commit only adds for hypervisors that have support for it.

- qemu
- cloud-hypervisor

Fixes: #1926

Signed-off-by: Carlos Venegas <jos.c.venegas.munoz@intel.com>

src/runtime/virtcontainers/acrn.go

@@ -679,6 +679,10 @@ func (a *Acrn) getPids() []int {
 	return []int{a.state.PID}
 }
 
+func (a *Acrn) getVirtioFsPid() *int {
+	return nil
+}
+
 func (a *Acrn) fromGrpc(ctx context.Context, hypervisorConfig *HypervisorConfig, j []byte) error {
 	return errors.New("acrn is not supported by VM cache")
 }

src/runtime/virtcontainers/clh.go

@@ -712,6 +712,10 @@ func (clh *cloudHypervisor) getPids() []int {
 	return pids
 }
 
+func (clh *cloudHypervisor) getVirtioFsPid() *int {
+	return &clh.state.VirtiofsdPID
+}
+
 func (clh *cloudHypervisor) addDevice(ctx context.Context, devInfo interface{}, devType deviceType) error {
 	span, _ := clh.trace(ctx, "addDevice")
 	defer span.End()

src/runtime/virtcontainers/fc.go

@@ -1181,6 +1181,10 @@ func (fc *firecracker) getPids() []int {
 	return []int{fc.info.PID}
 }
 
+func (fc *firecracker) getVirtioFsPid() *int {
+	return nil
+}
+
 func (fc *firecracker) fromGrpc(ctx context.Context, hypervisorConfig *HypervisorConfig, j []byte) error {
 	return errors.New("firecracker is not supported by VM cache")
 }

src/runtime/virtcontainers/hypervisor.go

@@ -818,6 +818,7 @@ type hypervisor interface {
 	// getPids returns a slice of hypervisor related process ids.
 	// The hypervisor pid must be put at index 0.
 	getPids() []int
+	getVirtioFsPid() *int
 	fromGrpc(ctx context.Context, hypervisorConfig *HypervisorConfig, j []byte) error
 	toGrpc(ctx context.Context) ([]byte, error)
 	check() error

src/runtime/virtcontainers/mock_hypervisor.go

@@ -109,6 +109,10 @@ func (m *mockHypervisor) getPids() []int {
 	return []int{m.mockPid}
 }
 
+func (m *mockHypervisor) getVirtioFsPid() *int {
+	return nil
+}
+
 func (m *mockHypervisor) fromGrpc(ctx context.Context, hypervisorConfig *HypervisorConfig, j []byte) error {
 	return errors.New("mockHypervisor is not supported by VM cache")
 }

src/runtime/virtcontainers/qemu.go

@@ -2329,6 +2329,10 @@ func (q *qemu) getPids() []int {
 	return pids
 }
 
+func (q *qemu) getVirtioFsPid() *int {
+	return &q.state.VirtiofsdPid
+}
+
 type qemuGrpc struct {
 	ID             string
 	QmpChannelpath string

src/runtime/virtcontainers/sandbox_metrics.go

@@ -16,8 +16,10 @@ import (
 
 const namespaceHypervisor = "kata_hypervisor"
 const namespaceKatashim = "kata_shim"
+const namespaceVirtiofsd = "kata_virtiofsd"
 
 var (
+	// hypervisor
 	hypervisorThreads = prometheus.NewGauge(prometheus.GaugeOpts{
 		Namespace: namespaceHypervisor,
 		Name:      "threads",
@@ -62,6 +64,7 @@ var (
 		Help:      "Open FDs for hypervisor.",
 	})
 
+	// agent
 	agentRPCDurationsHistogram = prometheus.NewHistogramVec(prometheus.HistogramOpts{
 		Namespace: namespaceKatashim,
 		Name:      "agent_rpc_durations_histogram_milliseconds",
@@ -70,16 +73,61 @@ var (
 	},
 		[]string{"action"},
 	)
+
+	// virtiofsd
+	virtiofsdThreads = prometheus.NewGauge(prometheus.GaugeOpts{
+		Namespace: namespaceVirtiofsd,
+		Name:      "threads",
+		Help:      "Virtiofsd process threads.",
+	})
+
+	virtiofsdProcStatus = prometheus.NewGaugeVec(prometheus.GaugeOpts{
+		Namespace: namespaceVirtiofsd,
+		Name:      "proc_status",
+		Help:      "Virtiofsd process status.",
+	},
+		[]string{"item"},
+	)
+
+	virtiofsdProcStat = prometheus.NewGaugeVec(prometheus.GaugeOpts{
+		Namespace: namespaceVirtiofsd,
+		Name:      "proc_stat",
+		Help:      "Virtiofsd process statistics.",
+	},
+		[]string{"item"},
+	)
+
+	virtiofsdIOStat = prometheus.NewGaugeVec(prometheus.GaugeOpts{
+		Namespace: namespaceVirtiofsd,
+		Name:      "io_stat",
+		Help:      "Process IO statistics.",
+	},
+		[]string{"item"},
+	)
+
+	virtiofsdOpenFDs = prometheus.NewGauge(prometheus.GaugeOpts{
+		Namespace: namespaceVirtiofsd,
+		Name:      "fds",
+		Help:      "Open FDs for virtiofsd.",
+	})
 )
 
 func RegisterMetrics() {
+	// hypervisor
 	prometheus.MustRegister(hypervisorThreads)
 	prometheus.MustRegister(hypervisorProcStatus)
 	prometheus.MustRegister(hypervisorProcStat)
 	prometheus.MustRegister(hypervisorNetdev)
 	prometheus.MustRegister(hypervisorIOStat)
 	prometheus.MustRegister(hypervisorOpenFDs)
+	// agent
 	prometheus.MustRegister(agentRPCDurationsHistogram)
+	// virtiofsd
+	prometheus.MustRegister(virtiofsdThreads)
+	prometheus.MustRegister(virtiofsdProcStatus)
+	prometheus.MustRegister(virtiofsdProcStat)
+	prometheus.MustRegister(virtiofsdIOStat)
+	prometheus.MustRegister(virtiofsdOpenFDs)
 }
 
 // UpdateRuntimeMetrics update shim/hypervisor's metrics
@@ -125,6 +173,48 @@ func (s *Sandbox) UpdateRuntimeMetrics() error {
 		mutils.SetGaugeVecProcIO(hypervisorIOStat, ioStat)
 	}
 
+	// virtiofs metrics
+	err = s.UpdateVirtiofsdMetrics()
+	if err != nil {
+		return err
+	}
+
+	return nil
+}
+
+func (s *Sandbox) UpdateVirtiofsdMetrics() error {
+	vfsPid := s.hypervisor.getVirtioFsPid()
+	if vfsPid == nil {
+		// virtiofsd is not mandatory for a VMM.
+		return nil
+	}
+
+	proc, err := procfs.NewProc(*vfsPid)
+	if err != nil {
+		return err
+	}
+
+	// process FDs
+	if fds, err := proc.FileDescriptorsLen(); err == nil {
+		virtiofsdOpenFDs.Set(float64(fds))
+	}
+
+	// process statistics
+	if procStat, err := proc.Stat(); err == nil {
+		virtiofsdThreads.Set(float64(procStat.NumThreads))
+		mutils.SetGaugeVecProcStat(virtiofsdProcStat, procStat)
+	}
+
+	// process status
+	if procStatus, err := proc.NewStatus(); err == nil {
+		mutils.SetGaugeVecProcStatus(virtiofsdProcStatus, procStatus)
+	}
+
+	// process IO statistics
+	if ioStat, err := proc.IO(); err == nil {
+		mutils.SetGaugeVecProcIO(virtiofsdIOStat, ioStat)
+	}
+
 	return nil
 }