dcgm-exporter

Tested component versions: DCGM 4.6.1, go-dcgm v1.4611.2, go-nvml v0.13.4-0, Go 1.27.1, and NVIDIA distroless CC v4.1.4.

Runtime and performance

• Release DCGM Exporter 4.8.4 with DCGM 4.6.1 and go-dcgm v1.4611.2.
• Add configurable DCGM field-watch retention limits through CLI, environment, and YAML configuration.
• Add the version 2 YAML schema with plural collections while preserving version 1 compatibility.
• Add configurable limits and request coalescing for concurrent /metrics scrapes.
• Reduce allocations during metric gathering, transformation, and rendering.
• Support metric configurations containing more than 128 DCGM fields by batching reads and splitting oversized field groups.
• Add optional GPU bind/unbind detection that resets DCGM and NVML, then rebuilds metrics for the new GPU topology (#706, reported by @varunrsekar; #716 by @dharmjit).
• Honor LD_LIBRARY_PATH when locating DCGM libraries.

Metrics and Kubernetes

• Add opt-in Go runtime, process, and HTTP handler metrics through --enable-exporter-metrics, DCGM_EXPORTER_ENABLE_EXPORTER_METRICS, or YAML configuration.
• Add opt-in GPU fabric health fields: DCGM_FI_DEV_FABRIC_HEALTH_SUMMARY and DCGM_FI_DEV_FABRIC_HEALTH_MASK.
• Add opt-in configured and enforced power-limit fields: DCGM_FI_DEV_POWER_MGMT_LIMIT and DCGM_FI_DEV_ENFORCED_POWER_LIMIT (#719 by @kumaraditya18027).
• Refresh the optional NVLink error-counter definitions.
• Remove retired PCIe throughput fields from the supplied metric configurations. Custom configurations that use them fail and point to DCGM_FI_PROF_PCIE_TX_BYTES and DCGM_FI_PROF_PCIE_RX_BYTES.
• Collect compute-instance-scoped DCGM fields from MIG compute instances and add the GPU_CI_ID label while preserving whole-GPU collection.
• Attribute DRA MIG metrics to pods when the MIG device is visible to DCGM at startup, and refresh allocation metadata when ResourceSlices change (#714, reported by @guptaNswati).
• Require an exporter restart after live Dynamic MIG creation or removal; changing the MIG layout while the exporter is running is not supported.
• Add the optional pod_uid label to DRA-attributed metrics (#738 by @git-jxj; fixes #718).
• Reuse the kubelet pod-resources gRPC connection to prevent memory growth (#705 by @AshwinRamaniPsg).
• Prefer NVML MIG profile names and make MIG profile lookup available outside Kubernetes (#713, reported by @rahulait).
• Deduplicate repeated job IDs within each HPC mapping file while preserving mappings across files.
• Warn when HPC job mapping entries do not match discovered MIG instances.

Helm, containers, and packages

• Include all supported metric preset CSV files in binary packages.
• Include third-party notices in archives, packages, and container images.
• Include the dynamic linker cache in the distroless image so DCGM libraries can be found at startup (#724 by @arielrahamim; fixes #612).
• Add Helm values for DaemonSet annotations (#723 by @ranyhb).
• Add Helm values for readiness and liveness probe timing and failure thresholds (#679 by @ChenZiHong-Gavin).
• Render debug dump compression explicitly, including when it is disabled.
• Document opt-in scheduling on nodes with the nvidia.com/gpu:NoSchedule taint (#733 by @KR-Ravindra).

Changelist

Runtime and configuration

• Release DCGM Exporter 4.8.3 with DCGM 4.6.0 and go-dcgm v1.4601.1 (@nccurry).
• Improve go-dcgm policy, resource, field, and version API handling.
• Add YAML configuration-file support.
• Expand CLI and environment configuration.
• Add configurable web read/write timeouts.
• Add per-watch polling intervals.
• Add combined device selectors.
• Add VSOCK remote hostengine connections.
• Use the local hostname for loopback remote hostengines (#697 by @skartikey).
• Improve hot reload and last-known-good configuration handling.
• Improve profiling and watcher recovery.
• Improve startup, skipped-field, MIG, and remote-hostengine diagnostics.
• Strengthen opt-in pprof and socket validation.

Metrics and Kubernetes

• Rename the Hostname label to hostname (#655 by @daphne2337).
• Add generic container-runtime label collection.
• Support pod-label collection on MIG devices (#660 by @Raid57).
• Allocate independent label maps for each entity (#663 by @rocker-zhang).
• Add DRA ResourceSlice v1 support (#654 by @adityasingh0510).
• Expand DRA v1 and MIG allocation handling (#664 by @guptaNswati).
• Bound pod-resources responses to prevent resource-exhaustion errors (#510 by @halcyon-r).
• Improve Kubernetes metadata diagnostics.
• Add cumulative XID error totals.
• Add cumulative clock-event totals.
• Add GPU-health severity and category metadata.
• Add Grace CPU serial labels.
• Improve metric help text and metadata consistency (@rvatkar, @dshaiknvidia).
• Correct NVLink bandwidth metrics from counters to gauges (#658 by @allenz92).
• Harden metric float parsing (#489 by @tongyangyeyue).
• Improve field-status handling, metric exposition, and label sanitization.
• Propagate HTTP request context through metric rendering (#687 by @randomizedcoder).

Helm, containers, and packages

• Add Helm support for configuration files.
• Add Helm image-digest configuration.
• Update ServiceMonitor timing defaults.
• Align RBAC with Kubernetes metadata and DRA features.
• Make service-account token mounting configurable (#665 by @75asu).
• Send package-managed systemd logs to journald (#671).
• Correct container image provenance metadata (#678 by @shravyavorugallu).
• Improve DCGM, NVPerf, and NVSDM library packaging (#704 by @daveoy, #703).
• Complete the Go 1.26.4 update (#641, #667).
• Remove the direct dependency on github.com/pkg/errors (#699 by @skartikey).
• Centralize release versions and consistency validation.

Testing and documentation

• Add a reusable E2E CLI and scenario framework.
• Expand static, host, container, and Kubernetes test coverage.
• Add repository-local agent-assisted development guidance.
• Add the llms.txt metrics and deployment authoring contract.
• Add Codex and Cursor project guidance.

• Update to DCGM 4.5.3 and DCGM Exporter 4.8.2.
• Improve GPU health metrics, including reporting GPU-wide health incidents such as fallen-off-bus XIDs.
• Make /debug/pprof profiling endpoints opt-in via --enable-pprof / DCGM_EXPORTER_ENABLE_PPROF.
• Add PodMapper informer caching for Kubernetes pod mapping (#626) (@jaeeyoungkim).
• Add per-process GPU metrics for time-sharing and MIG (#594) (@krystiancastai).
• Make Helm priorityClassName configurable with explicit defaults (#444) (@runzhliu).
• Add MIG device support for HPC job labels (#602) (@jay-mckay).
• Update go-dcgm field metadata handling, deprecated field alias resolution, health constants, policy registration handling, and version info APIs.
• Document IPv6 address formats for remote hostengine and metrics listen addresses.
• Refresh dependencies, container base images, Docker image references, Helm chart values, Kubernetes manifests, and tests for this release.

• Update to DCGM 4.5.2, latest Go 1.24, and base containers
• Fix distroless symlink issue
• Fix for parsing blank XIDs
• Fix for nvlink entities starting at offset 1