Skip to content

feat(nvidia_smi): add schema v13 support #2156

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
zhihonl wants to merge 2 commits into
base: main
Choose a base branch
from
Open

feat(nvidia_smi): add schema v13 support #2156

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
cf03992
feat(nvidia_smi): add schema v13 support
zhihonl Jun 12, 2026
290aa34
Merge branch 'main' into fix/nvidia-smi-v13-schema
zhihonl Jun 12, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
5 changes: 4 additions & 1 deletion plugins/inputs/nvidia_smi/nvidia_smi.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -24,6 +24,7 @@ import (
"github.com/aws/amazon-cloudwatch-agent/internal"
"github.com/aws/amazon-cloudwatch-agent/plugins/inputs/nvidia_smi/schema_v11"
"github.com/aws/amazon-cloudwatch-agent/plugins/inputs/nvidia_smi/schema_v12"
"github.com/aws/amazon-cloudwatch-agent/plugins/inputs/nvidia_smi/schema_v13"
)

//go:embed sample.conf
Expand Down Expand Up @@ -123,14 +124,16 @@ func (smi *NvidiaSMI) parse(acc telegraf.Accumulator, data []byte) error {
return schema_v11.Parse(acc, data)
case "v12":
return schema_v12.Parse(acc, data)
case "v13":
return schema_v13.Parse(acc, data)
}

smi.once.Do(func() {
smi.Log.Warnf(`Unknown schema version %q, using latest know schema for parsing.
Please report this as an issue to https://github.com/influxdata/telegraf together
with a sample output of 'nvidia_smi -q -x'!`, schema)
})
return schema_v12.Parse(acc, data)
return schema_v13.Parse(acc, data)
}

func init() {
Expand Down
53 changes: 53 additions & 0 deletions plugins/inputs/nvidia_smi/nvidia_smi_test.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -577,6 +577,59 @@ func TestGatherValidXML(t *testing.T) {
time.Unix(1689872450, 0)),
},
},
{
name: "RTX 3080 schema v13",
filename: "rtx-3080-v13.xml",
expected: []telegraf.Metric{
testutil.MustMetric(
"nvidia_smi",
map[string]string{
"arch": "Ampere",
"compute_mode": "Default",
"index": "0",
"name": "NVIDIA GeForce RTX 3080",
"pstate": "P2",
"uuid": "GPU-19d6d965-2acc-f646-00f8-4c76979aabb4",
},
map[string]interface{}{
"clocks_current_graphics": 1950,
"clocks_current_memory": 9251,
"clocks_current_sm": 1950,
"clocks_current_video": 1710,
"cuda_version": "13.2",
"display_active": "Disabled",
"display_mode": "Requested",
"driver_version": "595.58.03",
"ecc_errors_channel_repair_pending": "No",
"ecc_errors_tpc_repair_pending": "No",
"encoder_stats_average_fps": 0,
"encoder_stats_average_latency": 0,
"encoder_stats_session_count": 0,
"fan_speed": 0,
"fbc_stats_average_fps": 0,
"fbc_stats_average_latency": 0,
"fbc_stats_session_count": 0,
"memory_free": 660,
"memory_reserved": 397,
"memory_total": 10240,
"memory_used": 9184,
"pcie_link_gen_current": 4,
"pcie_link_width_current": 16,
"power_draw": 142.33,
"power_limit": 320.0,
"serial": "REDACTED",
"temperature_gpu": 27,
"utilization_decoder": 0,
"utilization_encoder": 0,
"utilization_gpu": 65,
"utilization_jpeg": 0,
"utilization_memory": 9,
"utilization_ofa": 0,
"vbios_version": "94.02.71.40.72",
},
time.Unix(1744280581, 0)),
},
},
}
for _, tt := range tests {
t.Run(tt.name, func(t *testing.T) {
Expand Down
157 changes: 157 additions & 0 deletions plugins/inputs/nvidia_smi/schema_v13/parser.go
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,157 @@
// Copyright Amazon.com, Inc. or its affiliates. All Rights Reserved.
// SPDX-License-Identifier: MIT

package schema_v13

import (
"encoding/xml"
"strconv"
"time"

"github.com/influxdata/telegraf"

"github.com/aws/amazon-cloudwatch-agent/plugins/inputs/nvidia_smi/common"
)

// Parse parses the XML-encoded data from nvidia-smi and adds measurements.
func Parse(acc telegraf.Accumulator, buf []byte) error {
var s smi
if err := xml.Unmarshal(buf, &s); err != nil {
return err
}

timestamp := time.Now()
if s.Timestamp != "" {
if t, err := time.ParseInLocation(time.ANSIC, s.Timestamp, time.Local); err == nil {
timestamp = t
}
}

for i := range s.Gpu {
gpu := &s.Gpu[i]

tags := map[string]string{
"index": strconv.Itoa(i),
}
fields := map[string]interface{}{}

common.SetTagIfUsed(tags, "pstate", gpu.PerformanceState)
common.SetTagIfUsed(tags, "name", gpu.ProductName)
common.SetTagIfUsed(tags, "arch", gpu.ProductArchitecture)
common.SetTagIfUsed(tags, "uuid", gpu.UUID)
common.SetTagIfUsed(tags, "compute_mode", gpu.ComputeMode)

common.SetIfUsed("str", fields, "driver_version", s.DriverVersion)
common.SetIfUsed("str", fields, "cuda_version", s.CudaVersion)
common.SetIfUsed("str", fields, "serial", gpu.Serial)
common.SetIfUsed("str", fields, "vbios_version", gpu.VbiosVersion)
common.SetIfUsed("str", fields, "display_active", gpu.DisplayActive)
common.SetIfUsed("str", fields, "display_mode", gpu.DisplayMode)
common.SetIfUsed("str", fields, "current_ecc", gpu.EccMode.CurrentEcc)
common.SetIfUsed("int", fields, "fan_speed", gpu.FanSpeed)
common.SetIfUsed("int", fields, "memory_total", gpu.FbMemoryUsage.Total)
common.SetIfUsed("int", fields, "memory_used", gpu.FbMemoryUsage.Used)
common.SetIfUsed("int", fields, "memory_free", gpu.FbMemoryUsage.Free)
common.SetIfUsed("int", fields, "memory_reserved", gpu.FbMemoryUsage.Reserved)
// v13 adds detailed ECC error breakdowns (volatile/aggregate, SRAM source attribution)
// that are absent from the v12 schema.
common.SetIfUsed("int", fields, "ecc_errors_volatile_dram_correctable", gpu.EccErrors.Volatile.DramCorrectable)
common.SetIfUsed("int", fields, "ecc_errors_volatile_dram_uncorrectable", gpu.EccErrors.Volatile.DramUncorrectable)
common.SetIfUsed("int", fields, "ecc_errors_volatile_sram_correctable", gpu.EccErrors.Volatile.SramCorrectable)
common.SetIfUsed("int", fields, "ecc_errors_volatile_sram_uncorrectable", gpu.EccErrors.Volatile.SramUncorrectable)
common.SetIfUsed("int", fields, "ecc_errors_volatile_sram_uncorrectable_parity", gpu.EccErrors.Volatile.SramUncorrectableParity)
common.SetIfUsed("int", fields, "ecc_errors_volatile_sram_uncorrectable_secded", gpu.EccErrors.Volatile.SramUncorrectableSecded)
common.SetIfUsed("int", fields, "ecc_errors_aggregate_dram_correctable", gpu.EccErrors.Aggregate.DramCorrectable)
common.SetIfUsed("int", fields, "ecc_errors_aggregate_dram_uncorrectable", gpu.EccErrors.Aggregate.DramUncorrectable)
common.SetIfUsed("int", fields, "ecc_errors_aggregate_sram_correctable", gpu.EccErrors.Aggregate.SramCorrectable)
common.SetIfUsed("int", fields, "ecc_errors_aggregate_sram_uncorrectable", gpu.EccErrors.Aggregate.SramUncorrectable)
common.SetIfUsed("int", fields, "ecc_errors_aggregate_sram_uncorrectable_parity", gpu.EccErrors.Aggregate.SramUncorrectableParity)
common.SetIfUsed("int", fields, "ecc_errors_aggregate_sram_uncorrectable_secded", gpu.EccErrors.Aggregate.SramUncorrectableSecded)
common.SetIfUsed("str", fields, "ecc_errors_aggregate_sram_threshold_exceeded", gpu.EccErrors.Aggregate.SramThresholdExceeded)
common.SetIfUsed("int", fields, "ecc_errors_aggregate_sram_uncorrectable_l2", gpu.EccErrors.AggregateUncorrectableSramSources.SramL2)
common.SetIfUsed("int", fields, "ecc_errors_aggregate_sram_uncorrectable_microcontroller",
gpu.EccErrors.AggregateUncorrectableSramSources.SramMicrocontroller)
common.SetIfUsed("int", fields, "ecc_errors_aggregate_sram_uncorrectable_other", gpu.EccErrors.AggregateUncorrectableSramSources.SramOther)
common.SetIfUsed("int", fields, "ecc_errors_aggregate_sram_uncorrectable_pcie", gpu.EccErrors.AggregateUncorrectableSramSources.SramPcie)
common.SetIfUsed("int", fields, "ecc_errors_aggregate_sram_uncorrectable_sm", gpu.EccErrors.AggregateUncorrectableSramSources.SramSm)
common.SetIfUsed("str", fields, "ecc_errors_channel_repair_pending", gpu.EccErrors.ChannelRepairPending)
common.SetIfUsed("str", fields, "ecc_errors_tpc_repair_pending", gpu.EccErrors.TpcRepairPending)
common.SetIfUsed("str", fields, "ecc_errors_unrepairable_memory", gpu.EccErrors.UnrepairableMemory)
common.SetIfUsed("int", fields, "retired_pages_multiple_single_bit", gpu.RetiredPages.MultipleSingleBitRetirement.RetiredCount)
common.SetIfUsed("int", fields, "retired_pages_double_bit", gpu.RetiredPages.DoubleBitRetirement.RetiredCount)
common.SetIfUsed("str", fields, "retired_pages_blacklist", gpu.RetiredPages.PendingBlacklist)
common.SetIfUsed("str", fields, "retired_pages_pending", gpu.RetiredPages.PendingRetirement)
// v13 renames the remapped-row fields; reference the new struct field names directly.
common.SetIfUsed("int", fields, "remapped_rows_correctable", gpu.RemappedRows.RemappedRowCorr)
common.SetIfUsed("int", fields, "remapped_rows_uncorrectable", gpu.RemappedRows.RemappedRowUnc)
common.SetIfUsed("str", fields, "remapped_rows_pending", gpu.RemappedRows.RemappedRowPending)
common.SetIfUsed("str", fields, "remapped_rows_failure", gpu.RemappedRows.RemappedRowFailure)
common.SetIfUsed("int", fields, "temperature_gpu", gpu.Temperature.GpuTemp)
common.SetIfUsed("int", fields, "utilization_gpu", gpu.Utilization.GpuUtil)
common.SetIfUsed("int", fields, "utilization_memory", gpu.Utilization.MemoryUtil)
common.SetIfUsed("int", fields, "utilization_encoder", gpu.Utilization.EncoderUtil)
common.SetIfUsed("int", fields, "utilization_decoder", gpu.Utilization.DecoderUtil)
common.SetIfUsed("int", fields, "utilization_jpeg", gpu.Utilization.JpegUtil)
common.SetIfUsed("int", fields, "utilization_ofa", gpu.Utilization.OfaUtil)
common.SetIfUsed("int", fields, "pcie_link_gen_current", gpu.Pci.PciGpuLinkInfo.PcieGen.CurrentLinkGen)
common.SetIfUsed("int", fields, "pcie_link_width_current", gpu.Pci.PciGpuLinkInfo.LinkWidths.CurrentLinkWidth)
common.SetIfUsed("int", fields, "encoder_stats_session_count", gpu.EncoderStats.SessionCount)
common.SetIfUsed("int", fields, "encoder_stats_average_fps", gpu.EncoderStats.AverageFps)
common.SetIfUsed("int", fields, "encoder_stats_average_latency", gpu.EncoderStats.AverageLatency)
common.SetIfUsed("int", fields, "fbc_stats_session_count", gpu.FbcStats.SessionCount)
common.SetIfUsed("int", fields, "fbc_stats_average_fps", gpu.FbcStats.AverageFps)
common.SetIfUsed("int", fields, "fbc_stats_average_latency", gpu.FbcStats.AverageLatency)
common.SetIfUsed("int", fields, "clocks_current_graphics", gpu.Clocks.GraphicsClock)
common.SetIfUsed("int", fields, "clocks_current_sm", gpu.Clocks.SmClock)
common.SetIfUsed("int", fields, "clocks_current_memory", gpu.Clocks.MemClock)
common.SetIfUsed("int", fields, "clocks_current_video", gpu.Clocks.VideoClock)
common.SetIfUsed("float", fields, "power_draw", gpu.PowerReadings.PowerDraw)
common.SetIfUsed("float", fields, "power_draw", gpu.PowerReadings.InstantPowerDraw)
common.SetIfUsed("float", fields, "power_limit", gpu.PowerReadings.PowerLimit)
common.SetIfUsed("float", fields, "power_draw", gpu.GpuPowerReadings.PowerDraw)
common.SetIfUsed("float", fields, "power_draw", gpu.GpuPowerReadings.InstantPowerDraw)
common.SetIfUsed("float", fields, "power_limit", gpu.GpuPowerReadings.CurrentPowerLimit)
common.SetIfUsed("float", fields, "power_limit", gpu.GpuPowerReadings.PowerLimit)
common.SetIfUsed("float", fields, "module_power_draw", gpu.ModulePowerReadings.PowerDraw)
common.SetIfUsed("float", fields, "module_power_draw", gpu.ModulePowerReadings.InstantPowerDraw)
acc.AddFields("nvidia_smi", fields, tags, timestamp)

for _, device := range gpu.MigDevices.MigDevice {
tags := map[string]string{}
common.SetTagIfUsed(tags, "index", device.Index)
common.SetTagIfUsed(tags, "gpu_index", device.GpuInstanceID)
common.SetTagIfUsed(tags, "compute_index", device.ComputeInstanceID)
common.SetTagIfUsed(tags, "pstate", gpu.PerformanceState)
common.SetTagIfUsed(tags, "name", gpu.ProductName)
common.SetTagIfUsed(tags, "arch", gpu.ProductArchitecture)
common.SetTagIfUsed(tags, "uuid", gpu.UUID)
common.SetTagIfUsed(tags, "compute_mode", gpu.ComputeMode)

fields := map[string]interface{}{}
common.SetIfUsed("int", fields, "sram_uncorrectable", device.EccErrorCount.VolatileCount.SramUncorrectable)
common.SetIfUsed("int", fields, "memory_fb_total", device.FbMemoryUsage.Total)
common.SetIfUsed("int", fields, "memory_fb_reserved", device.FbMemoryUsage.Reserved)
common.SetIfUsed("int", fields, "memory_fb_used", device.FbMemoryUsage.Used)
common.SetIfUsed("int", fields, "memory_fb_free", device.FbMemoryUsage.Free)
common.SetIfUsed("int", fields, "memory_bar1_total", device.Bar1MemoryUsage.Total)
common.SetIfUsed("int", fields, "memory_bar1_used", device.Bar1MemoryUsage.Used)
common.SetIfUsed("int", fields, "memory_bar1_free", device.Bar1MemoryUsage.Free)

acc.AddFields("nvidia_smi_mig", fields, tags, timestamp)
}

for _, process := range gpu.Processes.ProcessInfo {
tags := make(map[string]string, 2)
common.SetTagIfUsed(tags, "name", process.ProcessName)
common.SetTagIfUsed(tags, "type", process.Type)

fields := make(map[string]interface{}, 2)
common.SetIfUsed("int", fields, "pid", process.Pid)
common.SetIfUsed("int", fields, "used_memory", process.UsedMemory)

acc.AddFields("nvidia_smi_process", fields, tags, timestamp)
}
}

return nil
}
Loading
Loading