Semantic conventions for GPU metrics

Status: Development

GPU metrics hw.gpu.*

Graphics Processing Unit (discrete).

hw.type MUST be set to "gpu".

All GPU metrics may include the below attributes:

Attribute	Type	Description	Examples	Requirement Level	Stability
hw.id	string	An identifier for the hardware component, unique within the monitored host	win32battery_battery_testsysa33_1	Required
hw.driver_version	string	Driver version for the hardware component	10.2.1-3	Recommended
hw.firmware_version	string	Firmware version of the hardware component	2.0.1	Recommended
hw.model	string	Descriptive model name of the hardware component	PERC H740P; Intel(R) Core(TM) i7-10700K; Dell XPS 15 Battery	Recommended
hw.name	string	An easily-recognizable name for the hardware component	eth0	Recommended
hw.parent	string	Unique identifier of the parent component (typically the hw.id attribute of the enclosure, or disk controller)	dellStorage_perc_0	Recommended
hw.serial_number	string	Serial number of the hardware component	CNFCP0123456789	Recommended
hw.vendor	string	Vendor name of the hardware component	Dell; HP; Intel; AMD; LSI; Lenovo	Recommended

Metric: hw.errors (GPU)

This metric is recommended.

Number of errors encountered by the GPU.

When using this metric, the following attributes MUST be set:

• hw.type MUST be set to "gpu" to indicate that the errors are from a GPU.
• error.type SHOULD be set to one of the following values to indicate the type of error:
  • "corrected": Errors that were detected and corrected by the GPU.
  • "uncorrected": Errors that were detected but could not be corrected by the GPU.

Name	Instrument Type	Unit (UCUM)	Description	Stability	Entity Associations
hw.errors	Counter	{error}	Number of errors encountered by the component.

Attribute	Type	Description	Examples	Requirement Level	Stability
hw.id	string	An identifier for the hardware component, unique within the monitored host	win32battery_battery_testsysa33_1	Required
hw.type	string	Type of the component [1]	battery; cpu; disk_controller	Required
error.type	string	The type of error encountered by the component. [2]	uncorrected; zero_buffer_credit; crc; bad_sector	Conditionally Required if and only if an error has occurred
hw.name	string	An easily-recognizable name for the hardware component	eth0	Recommended
hw.parent	string	Unique identifier of the parent component (typically the hw.id attribute of the enclosure, or disk controller)	dellStorage_perc_0	Recommended
network.io.direction	string	Direction of network traffic for network errors. [3]	receive; transmit	Recommended

[1] hw.type: Describes the category of the hardware component for which hw.state is being reported. For example, hw.type=temperature along with hw.state=degraded would indicate that the temperature of the hardware component has been reported as degraded.

[2] error.type: The error.type SHOULD match the error code reported by the component, the canonical name of the error, or another low-cardinality error identifier. Instrumentations SHOULD document the list of errors they report.

[3] network.io.direction: This attribute SHOULD only be used when hw.type is set to "network" to indicate the direction of the error.

error.type has the following list of well-known values. If one of them applies, then the respective value MUST be used; otherwise, a custom value MAY be used.

Value	Description	Stability
_OTHER	A fallback error value to be used when the instrumentation doesn’t define a custom value.

hw.type has the following list of well-known values. If one of them applies, then the respective value MUST be used; otherwise, a custom value MAY be used.

Value	Description	Stability
battery	Battery
cpu	CPU
disk_controller	Disk controller
enclosure	Enclosure
fan	Fan
gpu	GPU
logical_disk	Logical disk
memory	Memory
network	Network
physical_disk	Physical disk
power_supply	Power supply
tape_drive	Tape drive
temperature	Temperature
voltage	Voltage

network.io.direction has the following list of well-known values. If one of them applies, then the respective value MUST be used; otherwise, a custom value MAY be used.

Value	Description	Stability
receive	receive
transmit	transmit

Metric: hw.gpu.io

This metric is recommended.

Name	Instrument Type	Unit (UCUM)	Description	Stability	Entity Associations
hw.gpu.io	Counter	By	Received and transmitted bytes by the GPU.

Attribute	Type	Description	Examples	Requirement Level	Stability
hw.id	string	An identifier for the hardware component, unique within the monitored host	win32battery_battery_testsysa33_1	Required
network.io.direction	string	The network IO operation direction.	receive; transmit	Required
hw.driver_version	string	Driver version for the hardware component	10.2.1-3	Recommended
hw.firmware_version	string	Firmware version of the hardware component	2.0.1	Recommended
hw.model	string	Descriptive model name of the hardware component	PERC H740P; Intel(R) Core(TM) i7-10700K; Dell XPS 15 Battery	Recommended
hw.name	string	An easily-recognizable name for the hardware component	eth0	Recommended
hw.parent	string	Unique identifier of the parent component (typically the hw.id attribute of the enclosure, or disk controller)	dellStorage_perc_0	Recommended
hw.serial_number	string	Serial number of the hardware component	CNFCP0123456789	Recommended
hw.vendor	string	Vendor name of the hardware component	Dell; HP; Intel; AMD; LSI; Lenovo	Recommended

network.io.direction has the following list of well-known values. If one of them applies, then the respective value MUST be used; otherwise, a custom value MAY be used.

Value	Description	Stability
receive	receive
transmit	transmit

Metric: hw.gpu.memory.limit

This metric is recommended.

Name	Instrument Type	Unit (UCUM)	Description	Stability	Entity Associations
hw.gpu.memory.limit	UpDownCounter	By	Size of the GPU memory.

Attribute	Type	Description	Examples	Requirement Level	Stability
hw.id	string	An identifier for the hardware component, unique within the monitored host	win32battery_battery_testsysa33_1	Required
hw.driver_version	string	Driver version for the hardware component	10.2.1-3	Recommended
hw.firmware_version	string	Firmware version of the hardware component	2.0.1	Recommended
hw.model	string	Descriptive model name of the hardware component	PERC H740P; Intel(R) Core(TM) i7-10700K; Dell XPS 15 Battery	Recommended
hw.name	string	An easily-recognizable name for the hardware component	eth0	Recommended
hw.parent	string	Unique identifier of the parent component (typically the hw.id attribute of the enclosure, or disk controller)	dellStorage_perc_0	Recommended
hw.serial_number	string	Serial number of the hardware component	CNFCP0123456789	Recommended
hw.vendor	string	Vendor name of the hardware component	Dell; HP; Intel; AMD; LSI; Lenovo	Recommended

Metric: hw.gpu.memory.utilization

This metric is recommended.

Name	Instrument Type	Unit (UCUM)	Description	Stability	Entity Associations
hw.gpu.memory.utilization	Gauge	1	Fraction of GPU memory used.

Attribute	Type	Description	Examples	Requirement Level	Stability
hw.id	string	An identifier for the hardware component, unique within the monitored host	win32battery_battery_testsysa33_1	Required
hw.driver_version	string	Driver version for the hardware component	10.2.1-3	Recommended
hw.firmware_version	string	Firmware version of the hardware component	2.0.1	Recommended
hw.model	string	Descriptive model name of the hardware component	PERC H740P; Intel(R) Core(TM) i7-10700K; Dell XPS 15 Battery	Recommended
hw.name	string	An easily-recognizable name for the hardware component	eth0	Recommended
hw.parent	string	Unique identifier of the parent component (typically the hw.id attribute of the enclosure, or disk controller)	dellStorage_perc_0	Recommended
hw.serial_number	string	Serial number of the hardware component	CNFCP0123456789	Recommended
hw.vendor	string	Vendor name of the hardware component	Dell; HP; Intel; AMD; LSI; Lenovo	Recommended

Metric: hw.gpu.memory.usage

This metric is recommended.

Name	Instrument Type	Unit (UCUM)	Description	Stability	Entity Associations
hw.gpu.memory.usage	UpDownCounter	By	GPU memory used.

Attribute	Type	Description	Examples	Requirement Level	Stability
hw.id	string	An identifier for the hardware component, unique within the monitored host	win32battery_battery_testsysa33_1	Required
hw.driver_version	string	Driver version for the hardware component	10.2.1-3	Recommended
hw.firmware_version	string	Firmware version of the hardware component	2.0.1	Recommended
hw.model	string	Descriptive model name of the hardware component	PERC H740P; Intel(R) Core(TM) i7-10700K; Dell XPS 15 Battery	Recommended
hw.name	string	An easily-recognizable name for the hardware component	eth0	Recommended
hw.parent	string	Unique identifier of the parent component (typically the hw.id attribute of the enclosure, or disk controller)	dellStorage_perc_0	Recommended
hw.serial_number	string	Serial number of the hardware component	CNFCP0123456789	Recommended
hw.vendor	string	Vendor name of the hardware component	Dell; HP; Intel; AMD; LSI; Lenovo	Recommended

Metric: hw.gpu.utilization

This metric is recommended.

Name	Instrument Type	Unit (UCUM)	Description	Stability	Entity Associations
hw.gpu.utilization	Gauge	1	Fraction of time spent in a specific task.

Attribute	Type	Description	Examples	Requirement Level	Stability
hw.id	string	An identifier for the hardware component, unique within the monitored host	win32battery_battery_testsysa33_1	Required
hw.driver_version	string	Driver version for the hardware component	10.2.1-3	Recommended
hw.firmware_version	string	Firmware version of the hardware component	2.0.1	Recommended
hw.gpu.task	string	Type of task the GPU is performing	decoder; encoder; general	Recommended
hw.model	string	Descriptive model name of the hardware component	PERC H740P; Intel(R) Core(TM) i7-10700K; Dell XPS 15 Battery	Recommended
hw.name	string	An easily-recognizable name for the hardware component	eth0	Recommended
hw.parent	string	Unique identifier of the parent component (typically the hw.id attribute of the enclosure, or disk controller)	dellStorage_perc_0	Recommended
hw.serial_number	string	Serial number of the hardware component	CNFCP0123456789	Recommended
hw.vendor	string	Vendor name of the hardware component	Dell; HP; Intel; AMD; LSI; Lenovo	Recommended

hw.gpu.task has the following list of well-known values. If one of them applies, then the respective value MUST be used; otherwise, a custom value MAY be used.

Value	Description	Stability
decoder	Decoder
encoder	Encoder
general	General

Metric: hw.status (GPU)

This metric is recommended.

Operational status: 1 (true) or 0 (false) for each of the possible states.

When using this metric for GPU status, the following attributes MUST be set:

• hw.type MUST be set to "gpu" to indicate that the status is for a GPU.
• hw.state MUST be set to one of the following values to indicate the GPU state:
  • "ok": The GPU is operating normally.
  • "degraded": The GPU is operating with reduced functionality or performance.
  • "failed": The GPU has failed and is not operational.
  • "predicted_failure": The GPU is currently operational but is predicted to fail soon.

Name	Instrument Type	Unit (UCUM)	Description	Stability	Entity Associations
hw.status	UpDownCounter	1	Operational status: 1 (true) or 0 (false) for each of the possible states. [1]

[1]: hw.status is currently specified as an UpDownCounter but would ideally be represented using a StateSet as defined in OpenMetrics. This semantic convention will be updated once StateSet is specified in OpenTelemetry. This planned change is not expected to have any consequence on the way users query their timeseries backend to retrieve the values of hw.status over time.

Attribute	Type	Description	Examples	Requirement Level	Stability
hw.id	string	An identifier for the hardware component, unique within the monitored host	win32battery_battery_testsysa33_1	Required
hw.state	string	The current state of the component	degraded; failed; needs_cleaning	Required
hw.type	string	Type of the component [1]	battery; cpu; disk_controller	Required
hw.name	string	An easily-recognizable name for the hardware component	eth0	Recommended
hw.parent	string	Unique identifier of the parent component (typically the hw.id attribute of the enclosure, or disk controller)	dellStorage_perc_0	Recommended

[1] hw.type: Describes the category of the hardware component for which hw.state is being reported. For example, hw.type=temperature along with hw.state=degraded would indicate that the temperature of the hardware component has been reported as degraded.

hw.state has the following list of well-known values. If one of them applies, then the respective value MUST be used; otherwise, a custom value MAY be used.

Value	Description	Stability
degraded	Degraded
failed	Failed
needs_cleaning	Needs Cleaning
ok	OK
predicted_failure	Predicted Failure

hw.type has the following list of well-known values. If one of them applies, then the respective value MUST be used; otherwise, a custom value MAY be used.

Value	Description	Stability
battery	Battery
cpu	CPU
disk_controller	Disk controller
enclosure	Enclosure
fan	Fan
gpu	GPU
logical_disk	Logical disk
memory	Memory
network	Network
physical_disk	Physical disk
power_supply	Power supply
tape_drive	Tape drive
temperature	Temperature
voltage	Voltage