fix: Remove confused metrics
The exporter presently has metrics that are nonsense for a given type of drive, and remain at zero due to their defaults. Change the behavior to NOT emit a metric if the underlying JSON field is not present. Future related work may include parsing the corresponding metrics for SATA/SAS SSDs (e.g. `smartctl_device_percentage_used` could derived from `SSD_Life_Left` on some drives). Metrics no longer exported for the wrong type of drive: - `smartctl_device_nvme_capacity_bytes` (NVME-specific) - `smartctl_device_available_spare` (NVME-specific, ATA possible) - `smartctl_device_available_spare_threshold` (NVME-specific, ATA possible) - `smartctl_device_critical_warning` (NVME-specific, ATA possible) - `smartctl_device_interface_speed` (ATA-specific) - `smartctl_device_media_errors` (NVME-specific, ATA possible) - `smartctl_device_num_err_log_entries` (NVME-specific, SCSI uses distinct metrics, ATA possible) - `smartctl_device_nvme_capacity_bytes` (NVME-specific) - `smartctl_device_percentage_used` (NVME-specific, ATA possible) Signed-off-by: Robin H. Johnson <rjohnson@coreweave.com>
This commit is contained in:
Robin H. Johnson
parent
558a760c14
commit
d90594ac23
68
smartctl.go
68
smartctl.go
|
|
@ -29,6 +29,9 @@ type SMARTDevice struct {
|
|||
serial string
|
||||
family string
|
||||
model string
|
||||
// These are used to select types of metrics.
|
||||
interface_ string
|
||||
protocol string
|
||||
}
|
||||
|
||||
// SMARTctl object
|
||||
|
|
@ -41,6 +44,15 @@ type SMARTctl struct {
|
|||
|
||||
// NewSMARTctl is smartctl constructor
|
||||
func NewSMARTctl(logger log.Logger, json gjson.Result, ch chan<- prometheus.Metric) SMARTctl {
|
||||
var model_name string
|
||||
if obj := json.Get("model_name"); obj.Exists() {
|
||||
model_name = obj.String()
|
||||
}
|
||||
// If the drive returns an empty model name, replace that with unknown.
|
||||
if model_name == "" {
|
||||
model_name = "unknown"
|
||||
}
|
||||
|
||||
return SMARTctl{
|
||||
ch: ch,
|
||||
json: json,
|
||||
|
|
@ -49,7 +61,9 @@ func NewSMARTctl(logger log.Logger, json gjson.Result, ch chan<- prometheus.Metr
|
|||
device: strings.TrimPrefix(strings.TrimSpace(json.Get("device.name").String()), "/dev/"),
|
||||
serial: strings.TrimSpace(json.Get("serial_number").String()),
|
||||
family: strings.TrimSpace(GetStringIfExists(json, "model_family", "unknown")),
|
||||
model: strings.TrimSpace(json.Get("model_name").String()),
|
||||
model: strings.TrimSpace(model_name),
|
||||
interface_: strings.TrimSpace(json.Get("device.type").String()),
|
||||
protocol: strings.TrimSpace(json.Get("device.protocol").String()),
|
||||
},
|
||||
}
|
||||
}
|
||||
|
|
@ -66,12 +80,15 @@ func (smart *SMARTctl) Collect() {
|
|||
smart.minePowerOnSeconds()
|
||||
smart.mineRotationRate()
|
||||
smart.mineTemperatures()
|
||||
smart.minePowerCycleCount()
|
||||
smart.minePowerCycleCount() // ATA/SATA, NVME, SCSI, SAS
|
||||
smart.mineDeviceSCTStatus()
|
||||
smart.mineDeviceStatistics()
|
||||
smart.mineDeviceErrorLog()
|
||||
smart.mineDeviceSelfTestLog()
|
||||
smart.mineDeviceERC()
|
||||
smart.mineSmartStatus()
|
||||
|
||||
if smart.device.interface_ == "nvme" {
|
||||
smart.mineNvmePercentageUsed()
|
||||
smart.mineNvmeAvailableSpare()
|
||||
smart.mineNvmeAvailableSpareThreshold()
|
||||
|
|
@ -80,10 +97,13 @@ func (smart *SMARTctl) Collect() {
|
|||
smart.mineNvmeNumErrLogEntries()
|
||||
smart.mineNvmeBytesRead()
|
||||
smart.mineNvmeBytesWritten()
|
||||
smart.mineSmartStatus()
|
||||
}
|
||||
// SCSI, SAS
|
||||
if smart.device.interface_ == "scsi" {
|
||||
smart.mineSCSIGrownDefectList()
|
||||
smart.mineSCSIErrorCounterLog()
|
||||
}
|
||||
}
|
||||
|
||||
func (smart *SMARTctl) mineExitStatus() {
|
||||
smart.ch <- prometheus.MustNewConstMetric(
|
||||
|
|
@ -95,14 +115,13 @@ func (smart *SMARTctl) mineExitStatus() {
|
|||
}
|
||||
|
||||
func (smart *SMARTctl) mineDevice() {
|
||||
device := smart.json.Get("device")
|
||||
smart.ch <- prometheus.MustNewConstMetric(
|
||||
metricDeviceModel,
|
||||
prometheus.GaugeValue,
|
||||
1,
|
||||
smart.device.device,
|
||||
device.Get("type").String(),
|
||||
device.Get("protocol").String(),
|
||||
smart.device.interface_,
|
||||
smart.device.protocol,
|
||||
smart.device.family,
|
||||
smart.device.model,
|
||||
smart.device.serial,
|
||||
|
|
@ -130,13 +149,16 @@ func (smart *SMARTctl) mineCapacity() {
|
|||
smart.json.Get("user_capacity.bytes").Float(),
|
||||
smart.device.device,
|
||||
)
|
||||
nvme_total_capacity := smart.json.Get("nvme_total_capacity")
|
||||
if nvme_total_capacity.Exists() {
|
||||
smart.ch <- prometheus.MustNewConstMetric(
|
||||
metricDeviceTotalCapacityBytes,
|
||||
prometheus.GaugeValue,
|
||||
smart.json.Get("nvme_total_capacity").Float(),
|
||||
nvme_total_capacity.Float(),
|
||||
smart.device.device,
|
||||
)
|
||||
}
|
||||
}
|
||||
|
||||
func (smart *SMARTctl) mineBlockSize() {
|
||||
for _, blockType := range []string{"logical", "physical"} {
|
||||
|
|
@ -152,8 +174,10 @@ func (smart *SMARTctl) mineBlockSize() {
|
|||
|
||||
func (smart *SMARTctl) mineInterfaceSpeed() {
|
||||
iSpeed := smart.json.Get("interface_speed")
|
||||
if iSpeed.Exists() {
|
||||
for _, speedType := range []string{"max", "current"} {
|
||||
tSpeed := iSpeed.Get(speedType)
|
||||
if tSpeed.Exists() {
|
||||
smart.ch <- prometheus.MustNewConstMetric(
|
||||
metricDeviceInterfaceSpeed,
|
||||
prometheus.GaugeValue,
|
||||
|
|
@ -163,6 +187,8 @@ func (smart *SMARTctl) mineInterfaceSpeed() {
|
|||
)
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
func (smart *SMARTctl) mineDeviceAttribute() {
|
||||
for _, attribute := range smart.json.Get("ata_smart_attributes.table").Array() {
|
||||
|
|
@ -200,6 +226,8 @@ func (smart *SMARTctl) mineDeviceAttribute() {
|
|||
|
||||
func (smart *SMARTctl) minePowerOnSeconds() {
|
||||
pot := smart.json.Get("power_on_time")
|
||||
// If the power_on_time is NOT present, do not report as 0.
|
||||
if pot.Exists() {
|
||||
smart.ch <- prometheus.MustNewConstMetric(
|
||||
metricDevicePowerOnSeconds,
|
||||
prometheus.CounterValue,
|
||||
|
|
@ -207,9 +235,12 @@ func (smart *SMARTctl) minePowerOnSeconds() {
|
|||
smart.device.device,
|
||||
)
|
||||
}
|
||||
}
|
||||
|
||||
func (smart *SMARTctl) mineRotationRate() {
|
||||
rRate := GetFloatIfExists(smart.json, "rotation_rate", 0)
|
||||
// TODO: what should be done if this is absent vs really zero (for
|
||||
// solid-state drives)?
|
||||
if rRate > 0 {
|
||||
smart.ch <- prometheus.MustNewConstMetric(
|
||||
metricDeviceRotationRate,
|
||||
|
|
@ -237,12 +268,17 @@ func (smart *SMARTctl) mineTemperatures() {
|
|||
}
|
||||
|
||||
func (smart *SMARTctl) minePowerCycleCount() {
|
||||
// ATA & NVME
|
||||
powerCycleCount := smart.json.Get("power_cycle_count")
|
||||
if powerCycleCount.Exists() {
|
||||
smart.ch <- prometheus.MustNewConstMetric(
|
||||
metricDevicePowerCycleCount,
|
||||
prometheus.CounterValue,
|
||||
smart.json.Get("power_cycle_count").Float(),
|
||||
powerCycleCount.Float(),
|
||||
smart.device.device,
|
||||
)
|
||||
return
|
||||
}
|
||||
}
|
||||
|
||||
func (smart *SMARTctl) mineDeviceSCTStatus() {
|
||||
|
|
@ -312,25 +348,33 @@ func (smart *SMARTctl) mineNvmeNumErrLogEntries() {
|
|||
}
|
||||
|
||||
func (smart *SMARTctl) mineNvmeBytesRead() {
|
||||
blockSize := smart.json.Get("logical_block_size").Float()
|
||||
blockSize := smart.json.Get("logical_block_size")
|
||||
data_units_read := smart.json.Get("nvme_smart_health_information_log.data_units_read")
|
||||
if !blockSize.Exists() || !data_units_read.Exists() {
|
||||
return
|
||||
}
|
||||
smart.ch <- prometheus.MustNewConstMetric(
|
||||
metricDeviceBytesRead,
|
||||
prometheus.CounterValue,
|
||||
// This value is reported in thousands (i.e., a value of 1 corresponds to 1000 units of 512 bytes written) and is rounded up.
|
||||
// When the LBA size is a value other than 512 bytes, the controller shall convert the amount of data written to 512 byte units.
|
||||
smart.json.Get("nvme_smart_health_information_log.data_units_read").Float()*1000.0*blockSize,
|
||||
data_units_read.Float()*1000.0*blockSize.Float(),
|
||||
smart.device.device,
|
||||
)
|
||||
}
|
||||
|
||||
func (smart *SMARTctl) mineNvmeBytesWritten() {
|
||||
blockSize := smart.json.Get("logical_block_size").Float()
|
||||
blockSize := smart.json.Get("logical_block_size")
|
||||
data_units_written := smart.json.Get("nvme_smart_health_information_log.data_units_written")
|
||||
if !blockSize.Exists() || !data_units_written.Exists() {
|
||||
return
|
||||
}
|
||||
smart.ch <- prometheus.MustNewConstMetric(
|
||||
metricDeviceBytesWritten,
|
||||
prometheus.CounterValue,
|
||||
// This value is reported in thousands (i.e., a value of 1 corresponds to 1000 units of 512 bytes written) and is rounded up.
|
||||
// When the LBA size is a value other than 512 bytes, the controller shall convert the amount of data written to 512 byte units.
|
||||
smart.json.Get("nvme_smart_health_information_log.data_units_written").Float()*1000.0*blockSize,
|
||||
data_units_written.Float()*1000.0*blockSize.Float(),
|
||||
smart.device.device,
|
||||
)
|
||||
}
|
||||
|
|
|
|||
Loading…
Reference in New Issue