DCGM-Exporter 4.0.3 (#454)

Update to DCGM 4.1.1
Add support for GPU sharing metrics in k8s (@pintohutch)

Author: glowkey

.devcontainer/Dockerfile

@@ -74,4 +74,4 @@ ENV PATH $PATH:/usr/local/go/bin
 ENV NVIDIA_DRIVER_CAPABILITIES=compute,utility,compat32
 # disable all constraints on the configurations required by NVIDIA container toolkit
 ENV NVIDIA_DISABLE_REQUIRE="true"
-ENV NVIDIA_VISIBLE_DEVICES=all
+ENV NVIDIA_VISIBLE_DEVICES=all

.devcontainer/devcontainer.json

@@ -17,5 +17,6 @@
     "--security-opt",
     "seccomp=unconfined",
     "--gpus=all"
-    ]
-}
+    ],
+  "postStartCommand": "docker run --privileged --rm tonistiigi/binfmt --install all"
+}

Makefile

@@ -30,9 +30,9 @@ DOCKERCMD      := docker --debug buildx build
 MODULE         := github.com/NVIDIA/dcgm-exporter
 
 .PHONY: all binary install check-format local
-all: update-version ubuntu22.04 ubi9
+all: ubuntu22.04 ubi9
 
-binary: update-version
+binary:
 	cd cmd/dcgm-exporter; $(GO) build -ldflags "-X main.BuildVersion=${DCGM_VERSION}-${VERSION}"
 
 test-main: generate
@@ -46,7 +46,7 @@ check-format:
 	test $$(gofmt -l pkg | tee /dev/stderr | wc -l) -eq 0
 	test $$(gofmt -l cmd | tee /dev/stderr | wc -l) -eq 0
 
-push: update-version
+push:
 	$(MAKE) ubuntu22.04 OUTPUT=type=registry
 	$(MAKE) ubi9 OUTPUT=type=registry
 
@@ -60,13 +60,13 @@ endif
 ubi%: DOCKERFILE = docker/Dockerfile.ubi
 ubi%: --docker-build-%
 	@
-ubi9: BASE_IMAGE = nvcr.io/nvidia/cuda:12.6.3-base-ubi9
+ubi9: BASE_IMAGE = nvcr.io/nvidia/cuda:12.8.0-base-ubi9
 ubi9: IMAGE_TAG = ubi9
 
 ubuntu%: DOCKERFILE = docker/Dockerfile.ubuntu
 ubuntu%: --docker-build-%
 	@
-ubuntu22.04: BASE_IMAGE = nvcr.io/nvidia/cuda:12.6.3-base-ubuntu22.04
+ubuntu22.04: BASE_IMAGE = nvcr.io/nvidia/cuda:12.8.0-base-ubuntu22.04
 ubuntu22.04: IMAGE_TAG = ubuntu22.04
 
 
@@ -76,6 +76,7 @@ ubuntu22.04: IMAGE_TAG = ubuntu22.04
 	$(DOCKERCMD) --pull \
 		--output $(OUTPUT) \
 		--progress=plain \
+		--no-cache \
 		--platform $(PLATFORMS) \
 		--build-arg BASEIMAGE="$(BASE_IMAGE)" \
 		--build-arg "GOLANG_VERSION=$(GOLANG_VERSION)" \

README.md

@@ -11,7 +11,7 @@ Official documentation for DCGM-Exporter can be found on [docs.nvidia.com](https
 To gather metrics on a GPU node, simply start the `dcgm-exporter` container:
 
 ```shell
-docker run -d --gpus all --cap-add SYS_ADMIN --rm -p 9400:9400 nvcr.io/nvidia/k8s/dcgm-exporter:4.0.0-4.0.0-ubuntu22.04
+docker run -d --gpus all --cap-add SYS_ADMIN --rm -p 9400:9400 nvcr.io/nvidia/k8s/dcgm-exporter:4.1.1-4.0.3-ubuntu22.04
 curl localhost:9400/metrics
 # HELP DCGM_FI_DEV_SM_CLOCK SM clock frequency (in MHz).
 # TYPE DCGM_FI_DEV_SM_CLOCK gauge

dcgm-exporter.yaml

@@ -18,23 +18,23 @@ metadata:
   name: "dcgm-exporter"
   labels:
     app.kubernetes.io/name: "dcgm-exporter"
-    app.kubernetes.io/version: "4.0.0"
+    app.kubernetes.io/version: "4.0.3"
 spec:
   updateStrategy:
     type: RollingUpdate
   selector:
     matchLabels:
       app.kubernetes.io/name: "dcgm-exporter"
-      app.kubernetes.io/version: "4.0.0"
+      app.kubernetes.io/version: "4.0.3"
   template:
     metadata:
       labels:
         app.kubernetes.io/name: "dcgm-exporter"
-        app.kubernetes.io/version: "4.0.0"
+        app.kubernetes.io/version: "4.0.3"
       name: "dcgm-exporter"
     spec:
       containers:
-      - image: "nvcr.io/nvidia/k8s/dcgm-exporter:4.0.0-4.0.0-ubuntu22.04"
+      - image: "nvcr.io/nvidia/k8s/dcgm-exporter:4.1.1-4.0.3-ubuntu22.04"
         env:
         - name: "DCGM_EXPORTER_LISTEN"
           value: ":9400"
@@ -66,11 +66,11 @@ metadata:
   name: "dcgm-exporter"
   labels:
     app.kubernetes.io/name: "dcgm-exporter"
-    app.kubernetes.io/version: "4.0.0"
+    app.kubernetes.io/version: "4.0.3"
 spec:
   selector:
     app.kubernetes.io/name: "dcgm-exporter"
-    app.kubernetes.io/version: "4.0.0"
+    app.kubernetes.io/version: "4.0.3"
   ports:
   - name: "metrics"
     port: 9400

deployment/Chart.yaml

@@ -1,9 +1,9 @@
 apiVersion: v2
 name: dcgm-exporter
 description: A Helm chart for DCGM exporter
-version: "4.0.0"
+version: "4.0.3"
 kubeVersion: ">= 1.19.0-0"
-appVersion: "4.0.0"
+appVersion: "4.0.3"
 sources:
 - https://github.com/nvidia/dcgm-exporter
 home: https://github.com/nvidia/dcgm-exporter/

deployment/values.yaml

@@ -17,7 +17,7 @@ image:
   pullPolicy: IfNotPresent
   # Image tag defaults to AppVersion, but you can use the tag key
   # for the image tag, e.g:
-  tag: 4.0.0-4.0.0-ubuntu22.04
+  tag: 4.1.1-4.0.3-ubuntu22.04
 
 # Change the following reference to "/etc/dcgm-exporter/default-counters.csv"
 # to stop profiling metrics from DCGM

docker/Dockerfile.ubi

@@ -1,4 +1,4 @@
-ARG BASEIMAGE=nvcr.io/nvidia/cuda:12.6.3-base-ubi9
+ARG BASEIMAGE=nvcr.io/nvidia/cuda:12.8.0-base-ubi9
 
 FROM --platform=$BUILDPLATFORM ubuntu:18.04 AS builder
 

docker/Dockerfile.ubuntu

@@ -1,4 +1,4 @@
-ARG BASEIMAGE=nvcr.io/nvidia/cuda:12.6.3-base-ubuntu22.04
+ARG BASEIMAGE=nvcr.io/nvidia/cuda:12.8.0-base-ubuntu22.04
 
 FROM --platform=$BUILDPLATFORM ubuntu:18.04 AS builder
 
@@ -72,7 +72,6 @@ COPY etc /etc/dcgm-exporter
 ENV DEBIAN_FRONTEND=noninteractive
 RUN echo "$TARGETARCH" && apt-get -qq update && apt-get -qq install -y --no-install-recommends \
     datacenter-gpu-manager-4-core libcap2-bin \
-    && apt-get -qq purge --autoremove -y openssl \
     && apt-get -qq -y clean \
     && apt-get -qq -y autoclean \
     && apt-get -qq autoremove -y \

docker/dcgm-exporter-entrypoint.sh

@@ -1,6 +1,7 @@
 #!/usr/bin/env bash
 set -euo pipefail
 
+ldconfig #Must refresh the cache to find libdcgm.so
 # We want to setcap only when the container is started with the right caps
 DCGM_EXPORTER=$(readlink -f $(which dcgm-exporter))
 if [ -z "$NO_SETCAP" ]; then