dcgm-exporter
Helm chart
Request a free trial
Contact our team to test out this Helm chart and related images for free. Please also indicate any other images you would like to evaluate.
Tag:
1
# Copyright (c) 2020, NVIDIA CORPORATION. All rights reserved.
2
#
3
# Licensed under the Apache License, Version 2.0 (the "License");
4
# you may not use this file except in compliance with the License.
5
# You may obtain a copy of the License at
6
#
7
# http://www.apache.org/licenses/LICENSE-2.0
8
#
9
# Unless required by applicable law or agreed to in writing, software
10
# distributed under the License is distributed on an "AS IS" BASIS,
11
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
12
# See the License for the specific language governing permissions and
13
# limitations under the License.
14
15
image:
16
repository: cgr.dev/scratch-images/test-tmp/dcgm-exporter
17
pullPolicy: IfNotPresent
18
# Image tag defaults to AppVersion, but you can use the tag key
19
# for the image tag, e.g:
20
tag: 4.6.0.4.8.3-r2@sha256:60219daea21f202548b9786fd571c99c75e936682d74606816c587d77848a244
21
# Optional digest pin. When set, the chart renders repository@digest and
22
# ignores tag for the image reference.
23
digest: ""
24
# Add dcgm-exporter command line arguments here. When empty, the exporter uses
25
# its built-in default collectors path.
26
arguments: []
27
# NOTE: in general, add any command line arguments to arguments above and they
28
# will be passed through.
29
# Use "-r", "<HOST>:<PORT>" or a DCGM connection string to connect to an already running hostengine.
30
# Connection string formats: "tcp://<HOST>:<PORT>", "unix:///<SOCKET_PATH>", or "vsock://<CID>:<PORT>".
31
# Example arguments: ["-r", "host123:5555"]
32
# Example VSOCK arguments: ["-r", "vsock://3:5555"]
33
# Use "-n" to remove the hostname tag from the output.
34
# Example arguments: ["-n"]
35
# Use "-d" to specify the devices to monitor. -d must be followed by a string
36
# in the following format: f, g[:numeric_range], i[:numeric_range], or
37
# g[:numeric_range]+i[:numeric_range]. i+g ordering is also accepted.
38
# The flex option f is the default and cannot be combined with g or i.
39
# Where a numeric range is something like 0-4 or 0,2,4, etc.
40
# Example arguments: ["-d", "f"] to monitor all GPUs or GPU instances, or
41
# ["-d", "g:0-3"] to monitor GPUs 0-3, or ["-d", "i"] to monitor GPU instances.
42
# Use "-m" to specify the namespace and name of a configmap containing
43
# the watched exporter fields.
44
# Example arguments: ["-m", "default:exporter-metrics-config-map"]
45
# If adding "--enable-pprof", also enable tlsServerConfig and/or basicAuth.
46
# Pprof requires a web config file so /debug/pprof/ is protected by
47
# exporter-toolkit authentication or TLS.
48
49
# Optional dcgm-exporter YAML configuration. YAML is read at exporter startup.
50
# Changes to the YAML file require restarting the exporter pod.
51
config:
52
enabled: false
53
# When true, the chart creates a ConfigMap from config.data. When false,
54
# config.name must reference an existing ConfigMap.
55
create: true
56
# Defaults to "<release-name>-dcgm-exporter-config" when empty.
57
name: ""
58
key: config.yaml
59
mountPath: /etc/dcgm-exporter/config.yaml
60
data: |
61
version: 1
62
metrics:
63
file: /etc/dcgm-exporter/default-counters.csv
64
collection:
65
interval: 30s
66
# Image pull secrets for container images
67
imagePullSecrets: []
68
# Overrides the chart's name
69
nameOverride: ""
70
# Overrides the chart's computed fullname
71
fullnameOverride: ""
72
# Overrides the deployment namespace
73
namespaceOverride: ""
74
# hostPID allows the DCGM-Exporter container to see processes on the host node
75
# Default: false
76
hostPID: false
77
# Defines the runtime class that will be used by the pod
78
runtimeClassName: ""
79
# Defines the priority class that will be used by the pod
80
priorityClassName: "system-node-critical"
81
# Enable or disable host network
82
hostNetwork: false
83
# Defines serviceAccount names for components.
84
serviceAccount:
85
# Specifies whether a service account should be created
86
create: true
87
# Annotations to add to the service account
88
annotations: {}
89
# The name of the service account to use.
90
# If not set and create is true, a name is generated using the fullname template
91
name:
92
# Whether to mount the SA token on the DaemonSet pod and the ServiceAccount.
93
# If not set, derived from kubernetes.enablePodLabels/enablePodUID with rbac.create, or kubernetesDRA.enabled.
94
# Setting this to false disables in-cluster Kubernetes API authentication for those features.
95
automountServiceAccountToken:
96
rollingUpdate:
97
# Specifies maximum number of DaemonSet pods that can be unavailable during the update
98
maxUnavailable: 1
99
# Specifies maximum number of nodes with an existing available DaemonSet pod that can have an updated DaemonSet pod during during an update
100
maxSurge: 0
101
# Labels to be added to dcgm-exporter pods
102
podLabels: {}
103
# Annotations to be added to dcgm-exporter pods
104
podAnnotations: {}
105
# Using this annotation which is required for prometheus scraping
106
# prometheus.io/scrape: "true"
107
# prometheus.io/port: "9400"
108
109
# The SecurityContext for the dcgm-exporter pods
110
podSecurityContext: {}
111
# fsGroup: 2000
112
113
# The SecurityContext for the dcgm-exporter containers
114
securityContext:
115
runAsNonRoot: false
116
runAsUser: 0
117
capabilities:
118
add: ["SYS_ADMIN"] # Required for profiling metrics (DCGM_FI_PROF_*)
119
drop: ["ALL"]
120
allowPrivilegeEscalation: false
121
# readOnlyRootFilesystem: true # Enable if tmpfs volumes are configured
122
# Note: For non-root without profiling metrics, use:
123
# runAsNonRoot: true, runAsUser: 1000, and remove SYS_ADMIN from capabilities.add
124
# Defines the dcgm-exporter service
125
service:
126
# When enabled, the helm chart will create service
127
enable: true
128
type: ClusterIP
129
# Accepts either "Cluster" or "Local", choose Local if you want to route internal traffic within the node only
130
internalTrafficPolicy: Cluster
131
clusterIP: ""
132
port: 9400
133
address: ":9400"
134
# HTTP server read timeout for inbound scrape requests
135
webReadTimeout: 10s
136
# HTTP server write timeout for generating and returning scrape responses
137
webWriteTimeout: 30s
138
# Annotations to add to the service
139
annotations: {}
140
# Allows to control pod resources
141
resources:
142
limits:
143
cpu: 200m
144
memory: 512Mi
145
requests:
146
cpu: 100m
147
memory: 128Mi
148
serviceMonitor:
149
apiVersion: "monitoring.coreos.com/v1"
150
enabled: true
151
interval: 30s
152
scrapeTimeout: 25s
153
honorLabels: false
154
additionalLabels: {}
155
#monitoring: prometheus
156
relabelings: []
157
# - sourceLabels: [__meta_kubernetes_pod_node_name]
158
# separator: ;
159
# regex: ^(.*)$
160
# targetLabel: nodename
161
# replacement: $1
162
# action: replace
163
metricRelabelings: []
164
# - action: replace
165
# regex: original_metrics_name
166
# replacement: new_metrics_name
167
# sourceLabels:
168
# - name
169
# targetLabel: __name__
170
nodeSelector: {}
171
#node: gpu
172
173
tolerations:
174
- key: node-role.kubernetes.io/control-plane
175
operator: Exists
176
effect: NoSchedule
177
#- operator: Exists
178
179
affinity: {}
180
#nodeAffinity:
181
# requiredDuringSchedulingIgnoredDuringExecution:
182
# nodeSelectorTerms:
183
# - matchExpressions:
184
# - key: nvidia-gpu
185
# operator: Exists
186
187
extraHostVolumes: []
188
#- name: host-binaries
189
# hostPath: /opt/bin
190
191
extraConfigMapVolumes:
192
- name: exporter-metrics-volume
193
configMap:
194
name: exporter-metrics-config-map
195
items:
196
- key: metrics
197
path: default-counters.csv
198
extraVolumeMounts:
199
- name: exporter-metrics-volume
200
mountPath: /etc/dcgm-exporter/default-counters.csv
201
subPath: default-counters.csv
202
extraEnv: []
203
# Path to the kubelet socket for /pod-resources
204
kubeletPath: "/var/lib/kubelet/pod-resources"
205
# HTTPS configuration
206
tlsServerConfig:
207
# Enable or disable HTTPS configuration
208
enabled: false
209
# Use autogenerated self-signed TLS certificates. Not recommended for production environments.
210
autoGenerated: true
211
# Existing secret containing your own server key and certificate
212
existingSecret: ""
213
# Certificate file name
214
certFilename: "tls.crt"
215
# Key file name
216
keyFilename: "tls.key"
217
# CA certificate file name
218
caFilename: "ca.crt"
219
# Server policy for client authentication. Maps to ClientAuth Policies.
220
# For more detail on clientAuth options:
221
# https://golang.org/pkg/crypto/tls/#ClientAuthType
222
#
223
# NOTE: If you want to enable client authentication, you need to use
224
# RequireAndVerifyClientCert. Other values are insecure.
225
clientAuthType: ""
226
# TLS Key for HTTPS - ignored if existingSecret is provided
227
key: ""
228
# TLS Certificate for HTTPS - ignored if existingSecret is provided
229
cert: ""
230
# CA Certificate for HTTPS - ignored if existingSecret is provided
231
ca: ""
232
basicAuth:
233
#Object containing <user>:<passwords> key-value pairs for each user that will have access via basic authentication
234
users: {}
235
# Kubernetes integration settings
236
kubernetes:
237
# Enable Kubernetes pod labels in metrics
238
# When enabled, metrics will include labels from the pods that are using the GPUs
239
# This requires cluster-level read permissions to pods
240
enablePodLabels: false
241
# Enable Kubernetes pod UID in metrics
242
# When enabled, metrics will include the pod UID as an attribute for the pods that are using the GPUs
243
# This requires cluster-level read permissions to pods
244
enablePodUID: false
245
# Pod label filtering configuration
246
# Filter which pod labels are included in metrics using regex patterns
247
# Empty list means all labels are included (default behavior)
248
# Examples:
249
# podLabelAllowlistRegex:
250
# - "^app$" # Exact match for "app" label
251
# - "^app\\.kubernetes\\.io/.*" # All labels starting with app.kubernetes.io/
252
# - "^(tier|environment|version)$" # Match tier, environment, or version labels
253
podLabelAllowlistRegex: []
254
# RBAC settings for Kubernetes integration
255
rbac:
256
# Automatically creates ClusterRole and ClusterRoleBinding for pod access when enablePodLabels or enablePodUID is true
257
# Set to false if you want to manage RBAC resources manually
258
create: true
259
# Runtime object debug dump configuration for troubleshooting
260
debugDump:
261
# Enable file-based debugging dumps for troubleshooting
262
enabled: false
263
# Directory to store debug dump files
264
directory: "/tmp/dcgm-exporter-debug"
265
# Retention period for debug dump files in hours (0 = no cleanup)
266
retention: 24
267
# Use gzip compression for debug dump files
268
compression: true
269
# Enable DRA to capture metrics associated with GPUs managed by NVIDIA k8s-dra-driver (https://github.com/NVIDIA/k8s-dra-driver-gpu)
270
# default: false
271
kubernetesDRA:
272
enabled: false
273
# Customized list of metrics to emit. Expected to be in the same format (CSV) as the default list.
274
# Must be the complete list and is not additive. If unset, the default list will take effect.
275
# customMetrics: |
276
# Format
277
# If line starts with a '#' it is considered a comment
278
# DCGM FIELD, Prometheus metric type, help message
279
280
# Clocks
281
# DCGM_FI_DEV_SM_CLOCK, gauge, SM clock frequency (in MHz).
282
# DCGM_FI_DEV_MEM_CLOCK, gauge, Memory clock frequency (in MHz).
283
284
# Temperature
285
# DCGM_FI_DEV_MEMORY_TEMP, gauge, Memory temperature (in C).
286
# DCGM_FI_DEV_GPU_TEMP, gauge, GPU temperature (in C).
287
288
# Power
289
# DCGM_FI_DEV_POWER_USAGE, gauge, Power draw (in W).
290
# DCGM_FI_DEV_TOTAL_ENERGY_CONSUMPTION, counter, Total energy consumption since boot (in mJ).
291
292
# PCIE
293
# DCGM_FI_DEV_PCIE_TX_THROUGHPUT, counter, Total number of bytes transmitted through PCIe TX (in KB) via NVML.
294
# DCGM_FI_DEV_PCIE_RX_THROUGHPUT, counter, Total number of bytes received through PCIe RX (in KB) via NVML.
295
# DCGM_FI_DEV_PCIE_REPLAY_COUNTER, counter, Total number of PCIe retries.
296
297
# Utilization (the sample period varies depending on the product)
298
# DCGM_FI_DEV_GPU_UTIL, gauge, GPU utilization (in %).
299
# DCGM_FI_DEV_MEM_COPY_UTIL, gauge, Memory utilization (in %).
300
# DCGM_FI_DEV_ENC_UTIL, gauge, Encoder utilization (in %).
301
# DCGM_FI_DEV_DEC_UTIL , gauge, Decoder utilization (in %).
302
303
# Errors and violations
304
# DCGM_FI_DEV_XID_ERRORS, gauge, Value of the last XID error encountered.
305
# DCGM_EXP_CLOCK_EVENTS_TOTAL, counter, cumulative clock events observed since exporter start (edge-counted)
306
# DCGM_EXP_XID_ERRORS_TOTAL, counter, cumulative XID errors observed since exporter start
307
# DCGM_FI_DEV_POWER_VIOLATION, counter, Throttling duration due to power constraints (in ns).
308
# DCGM_FI_DEV_THERMAL_VIOLATION, counter, Throttling duration due to thermal constraints (in ns).
309
# DCGM_FI_DEV_SYNC_BOOST_VIOLATION, counter, Throttling duration due to sync-boost constraints (in ns).
310
# DCGM_FI_DEV_BOARD_LIMIT_VIOLATION, counter, Throttling duration due to board limit constraints (in ns).
311
# DCGM_FI_DEV_LOW_UTIL_VIOLATION, counter, Throttling duration due to low utilization (in ns).
312
# DCGM_FI_DEV_RELIABILITY_VIOLATION, counter, Throttling duration due to reliability constraints (in ns).
313
314
# Memory usage
315
# DCGM_FI_DEV_FB_FREE, gauge, Framebuffer memory free (in MiB).
316
# DCGM_FI_DEV_FB_USED, gauge, Framebuffer memory used (in MiB).
317
318
# ECC
319
# DCGM_FI_DEV_ECC_SBE_VOL_TOTAL, counter, Total number of single-bit volatile ECC errors.
320
# DCGM_FI_DEV_ECC_DBE_VOL_TOTAL, counter, Total number of double-bit volatile ECC errors.
321
# DCGM_FI_DEV_ECC_SBE_AGG_TOTAL, counter, Total number of single-bit persistent ECC errors.
322
# DCGM_FI_DEV_ECC_DBE_AGG_TOTAL, counter, Total number of double-bit persistent ECC errors.
323
324
# Retired pages
325
# DCGM_FI_DEV_RETIRED_SBE, counter, Total number of retired pages due to single-bit errors.
326
# DCGM_FI_DEV_RETIRED_DBE, counter, Total number of retired pages due to double-bit errors.
327
# DCGM_FI_DEV_RETIRED_PENDING, counter, Total number of pages pending retirement.
328
329
# NVLink
330
# DCGM_FI_DEV_NVLINK_CRC_FLIT_ERROR_COUNT_TOTAL, counter, Total number of NVLink flow-control CRC errors.
331
# DCGM_FI_DEV_NVLINK_CRC_DATA_ERROR_COUNT_TOTAL, counter, Total number of NVLink data CRC errors.
332
# DCGM_FI_DEV_NVLINK_REPLAY_ERROR_COUNT_TOTAL, counter, Total number of NVLink retries.
333
# DCGM_FI_DEV_NVLINK_RECOVERY_ERROR_COUNT_TOTAL, counter, Total number of NVLink recovery errors.
334
# DCGM_FI_DEV_NVLINK_BANDWIDTH_TOTAL, gauge, Total number of NVLink bandwidth counters for all lanes.
335
# DCGM_FI_DEV_NVLINK_BANDWIDTH_L0, counter, The number of bytes of active NVLink rx or tx data including both header and payload.
336
337
# VGPU License status
338
# DCGM_FI_DEV_VGPU_LICENSE_STATUS, gauge, vGPU License status
339
340
# Remapped rows
341
# DCGM_FI_DEV_UNCORRECTABLE_REMAPPED_ROWS, counter, Number of remapped rows for uncorrectable errors
342
# DCGM_FI_DEV_CORRECTABLE_REMAPPED_ROWS, counter, Number of remapped rows for correctable errors
343
# DCGM_FI_DEV_ROW_REMAP_FAILURE, gauge, Whether remapping of rows has failed
344
345
# DCP metrics
346
# DCGM_FI_PROF_GR_ENGINE_ACTIVE, gauge, Ratio of time the graphics engine is active.
347
# DCGM_FI_PROF_SM_ACTIVE, gauge, The ratio of cycles an SM has at least 1 warp assigned.
348
# DCGM_FI_PROF_SM_OCCUPANCY, gauge, The ratio of number of warps resident on an SM.
349
# DCGM_FI_PROF_PIPE_TENSOR_ACTIVE, gauge, Ratio of cycles the tensor (HMMA) pipe is active.
350
# DCGM_FI_PROF_DRAM_ACTIVE, gauge, Ratio of cycles the device memory interface is active sending or receiving data.
351
# DCGM_FI_PROF_PIPE_FP64_ACTIVE, gauge, Ratio of cycles the fp64 pipes are active.
352
# DCGM_FI_PROF_PIPE_FP32_ACTIVE, gauge, Ratio of cycles the fp32 pipes are active.
353
# DCGM_FI_PROF_PIPE_FP16_ACTIVE, gauge, Ratio of cycles the fp16 pipes are active.
354
# DCGM_FI_PROF_PCIE_TX_BYTES, gauge, The rate of data transmitted over the PCIe bus - including both protocol headers and data payloads - in bytes per second.
355
# DCGM_FI_PROF_PCIE_RX_BYTES, gauge, The rate of data received over the PCIe bus - including both protocol headers and data payloads - in bytes per second.
356
# DCGM_FI_PROF_SM_CYCLES_ELAPSED_TOTAL, counter, Total elapsed SM cycles.
357
# DCGM_FI_PROF_SM_CYCLES_ACTIVE_TOTAL, counter, Total SM cycles with active warps.
358
# DCGM_FI_PROF_MMA_CYCLES_ACTIVE_TOTAL, counter, Total MMA tensor cycles active.
359
# DCGM_FI_PROF_DMMA_CYCLES_ACTIVE_TOTAL, counter, Total DMMA tensor cycles active.
360
# DCGM_FI_PROF_HMMA_CYCLES_ACTIVE_TOTAL, counter, Total HMMA tensor cycles active.
361
# DCGM_FI_PROF_IMMA_CYCLES_ACTIVE_TOTAL, counter, Total IMMA tensor cycles active.
362
# DCGM_FI_PROF_DFMA_CYCLES_ACTIVE_TOTAL, counter, Total DFMA tensor cycles active.
363
# DCGM_FI_PROF_PCIE_TX_BYTES_TOTAL, counter, Total PCIe transmitted bytes.
364
# DCGM_FI_PROF_PCIE_RX_BYTES_TOTAL, counter, Total PCIe received bytes.
365
# DCGM_FI_PROF_INT_CYCLES_ACTIVE_TOTAL, counter, Total integer pipe cycles active.
366
# DCGM_FI_PROF_FP64_CYCLES_ACTIVE_TOTAL, counter, Total FP64 pipe cycles active.
367
# DCGM_FI_PROF_FP32_CYCLES_ACTIVE_TOTAL, counter, Total FP32 pipe cycles active.
368
# DCGM_FI_PROF_FP16_CYCLES_ACTIVE_TOTAL, counter, Total FP16 pipe cycles active.
369
livenessProbe:
370
initialDelaySeconds: 45
371
periodSeconds: 5
372
readinessProbe:
373
initialDelaySeconds: 45
374
The trusted source for open source
Talk to an expertProduct
Customers
© 2026 Chainguard, Inc. All Rights Reserved.
Chainguard® and the Chainguard logo are registered trademarks of Chainguard, Inc. in the United States and/or other countries.
The other respective trademarks mentioned on this page are owned by the respective companies and use of them does not imply any affiliation or endorsement.
Chainguard® and the Chainguard logo are registered trademarks of Chainguard, Inc. in the United States and/or other countries.
The other respective trademarks mentioned on this page are owned by the respective companies and use of them does not imply any affiliation or endorsement.