[EBPF-917] gpu: add initial CI job with GPUs

[gjulianm]

What does this PR do?

This PR enables tests on GPU runners, for improved integration testing and as an intermediate step between mocked unit tests and full e2e tests.

The new job only runs tests in two packages, specific for integration tests that will only run when there is a GPU available. These tests will be ignored in the rest of the test jobs.

To avoid high cost and usage of GPU capacity, it's configured to not run in all main pipelines, only on scheduled ones. On PRs, it will only run if GPU-related code is changed.

This PR includes some basic tests that use real data from NVML, the test set will later be expanded.

Motivation

Improve testing capabilities for GPUm.

Describe how you validated your changes

CI passing.

Additional Notes

[agent-platform-auto-pr[bot]]

Gitlab CI Configuration Changes

Added Jobs

.on_gpu_integration_tests

.on_gpu_integration_tests:
- if: $DDR_WORKFLOW_ID != null && $CI_COMMIT_BRANCH == "main"
  when: always
- if: $CI_COMMIT_BRANCH =~ /^mq-working-branch-/
  when: never
- allow_failure: true
  when: manual
- changes:
    compare_to: $COMPARE_TO_BRANCH
    paths:
    - pkg/gpu/**/*
    - pkg/collector/corechecks/gpu/**/*
    - comp/core/workloadmeta/collectors/internal/nvml/**/*
    - .gitlab/source_test/linux.yml
  when: on_success

tests_gpu

tests_gpu:
  after_script:
  - $CI_PROJECT_DIR/tools/ci/junit_upload.sh
  - CODECOV_TOKEN=$($CI_PROJECT_DIR/tools/ci/fetch_secret.sh $CODECOV token) || exit
    $?; export CODECOV_TOKEN
  - dda inv -- -e coverage.upload-to-codecov $COVERAGE_CACHE_FLAG || true
  allow_failure: true
  artifacts:
    expire_in: 2 weeks
    paths:
    - $TEST_OUTPUT_FILE
    - junit-*.tgz
    reports:
      annotations:
      - $EXTERNAL_LINKS_PATH
      junit:
      - '**/junit-out-*.xml'
    when: always
  before_script:
  - mkdir -p $GOPATH/pkg/mod/cache && tar xJf modcache.tar.xz -C $GOPATH/pkg/mod/cache
  - rm -f modcache.tar.xz
  cache:
  - key:
      files:
      - .bazelversion
      prefix: bazelversion-$CI_RUNNER_DESCRIPTION
    paths:
    - .cache/bazelisk
    - .cache/bazel/*/install
    policy: pull-push
    when: on_success
  - key: bazel-$CI_RUNNER_DESCRIPTION
    paths:
    - .cache/bazel/*/cache
    - .cache/bazel/*/*/external
    - .cache/pip
    policy: pull-push
    when: on_success
  image: registry.ddbuild.io/ci/datadog-agent-buildimages/linux$CI_IMAGE_LINUX_SUFFIX:$CI_IMAGE_LINUX
  needs:
  - go_deps
  - go_tools_deps
  rules:
  - if: $DDR_WORKFLOW_ID != null && $CI_COMMIT_BRANCH == "main"
    when: always
  - if: $CI_COMMIT_BRANCH =~ /^mq-working-branch-/
    when: never
  - allow_failure: true
    when: manual
  - changes:
      compare_to: $COMPARE_TO_BRANCH
      paths:
      - pkg/gpu/**/*
      - pkg/collector/corechecks/gpu/**/*
      - comp/core/workloadmeta/collectors/internal/nvml/**/*
      - .gitlab/source_test/linux.yml
    when: on_success
  - if: $RUN_UNIT_TESTS == "off"
    when: never
  script:
  - mkdir -p $GOPATH/pkg/mod/cache && tar xJf modcache_tools.tar.xz -C $GOPATH/pkg/mod/cache
  - rm -f modcache_tools.tar.xz
  - dda inv -- -e install-tools
  - dda inv -- -e gitlab.generate-ci-visibility-links --output=$EXTERNAL_LINKS_PATH
  - FAST_TESTS_FLAG=""
  - if [[ "$FAST_TESTS" == "true" ]]; then FAST_TESTS_FLAG="--only-impacted-packages";
    fi
  - dda inv -- -e agent.build
  - dda inv -- -e test $FLAVORS --race --profile --rerun-fails=2 --coverage --cpus
    $KUBERNETES_CPU_REQUEST $EXTRA_OPTS --result-json $TEST_OUTPUT_FILE --junit-tar
    "junit-${CI_JOB_NAME}.tgz" --build-stdlib $FAST_TESTS_FLAG --test-washer
  stage: source_test
  tags:
  - gpu:a10-amd64
  variables:
    BAZELISK_HOME: $XDG_CACHE_HOME/bazelisk
    EXTERNAL_LINKS_PATH: external_links_$CI_JOB_ID.json
    EXTRA_OPTS: --targets=./pkg/gpu/integrationtests,./pkg/collector/corechecks/gpu/integrationtests
    FLAKY_PATTERNS_CONFIG: $CI_PROJECT_DIR/flaky-patterns-runtime.yaml
    FLAVORS: ''
    KUBERNETES_CPU_REQUEST: 4
    KUBERNETES_MEMORY_LIMIT: 8Gi
    KUBERNETES_MEMORY_REQUEST: 8Gi
    TEST_OUTPUT_FILE: test_output.json
    XDG_CACHE_HOME: $CI_PROJECT_DIR/.cache

Changes Summary

Removed	Modified	Added	Renamed
0	0	2	0

:information_source: Diff available in the job log.

[agent-platform-auto-pr[bot]]

Static quality checks

✅ Please find below the results from static quality gates Comparison made with ancestor 69a72e41601e3d22eba91dae728760800a69c38d

Successful checks

Info

	Quality gate	Delta	On disk size (MiB)	Delta	On wire size (MiB)
✅	agent_deb_amd64	$${0}$$	$${705.4}$$ < $${707.13}$$	$${-0.01}$$	$${172.97}$$ < $${174.16}$$
✅	agent_deb_amd64_fips	$${0}$$	$${701.63}$$ < $${702.86}$$	$${-0.01}$$	$${172.22}$$ < $${173.5}$$
✅	agent_heroku_amd64	$${0}$$	$${327.71}$$ < $${327.92}$$	$${+0}$$	$${87.17}$$ < $${88.21}$$
✅	agent_msi	$${0}$$	$${571.1}$$ < $${982.08}$$	$${-0}$$	$${142.65}$$ < $${143.49}$$
✅	agent_rpm_amd64	$${0}$$	$${705.39}$$ < $${707.12}$$	$${-0.03}$$	$${176.1}$$ < $${177.29}$$
✅	agent_rpm_amd64_fips	$${0}$$	$${701.61}$$ < $${702.85}$$	$${-0.02}$$	$${175.07}$$ < $${176.32}$$
✅	agent_rpm_arm64	$${0}$$	$${687.26}$$ < $${691.67}$$	$${+0.01}$$	$${159.05}$$ < $${160.7}$$
✅	agent_rpm_arm64_fips	$${0}$$	$${684.32}$$ < $${687.54}$$	$${-0.01}$$	$${158.8}$$ < $${160.11}$$
✅	agent_suse_amd64	$${0}$$	$${705.39}$$ < $${707.12}$$	$${-0.03}$$	$${176.1}$$ < $${177.29}$$
✅	agent_suse_amd64_fips	$${0}$$	$${701.61}$$ < $${702.85}$$	$${-0.02}$$	$${175.07}$$ < $${176.32}$$
✅	agent_suse_arm64	$${0}$$	$${687.26}$$ < $${691.67}$$	$${+0.01}$$	$${159.05}$$ < $${160.7}$$
✅	agent_suse_arm64_fips	$${0}$$	$${684.32}$$ < $${687.54}$$	$${-0.01}$$	$${158.8}$$ < $${160.11}$$
✅	docker_agent_amd64	$${-0}$$	$${767.19}$$ < $${769.38}$$	$${-0}$$	$${260.51}$$ < $${262.06}$$
✅	docker_agent_arm64	$${+0}$$	$${773.63}$$ < $${778.35}$$	$${+0.01}$$	$${249.62}$$ < $${251.72}$$
✅	docker_agent_jmx_amd64	$${+0}$$	$${958.06}$$ < $${960.26}$$	$${+0}$$	$${329.15}$$ < $${330.69}$$
✅	docker_agent_jmx_arm64	$${+0}$$	$${953.22}$$ < $${957.95}$$	$${+0.01}$$	$${314.25}$$ < $${316.35}$$
✅	docker_cluster_agent_amd64	$${+0}$$	$${180.33}$$ < $${181.08}$$	$${+0}$$	$${63.71}$$ < $${64.49}$$
✅	docker_cluster_agent_arm64	$${0}$$	$${196.19}$$ < $${198.49}$$	$${+0}$$	$${60.0}$$ < $${61.17}$$
✅	docker_cws_instrumentation_amd64	$${0}$$	$${7.13}$$ < $${7.18}$$	$${+0}$$	$${2.99}$$ < $${3.33}$$
✅	docker_cws_instrumentation_arm64	$${0}$$	$${6.69}$$ < $${6.92}$$	$${+0}$$	$${2.73}$$ < $${3.09}$$
✅	docker_dogstatsd_amd64	$${-0}$$	$${38.76}$$ < $${39.38}$$	$${+0}$$	$${15.0}$$ < $${15.82}$$
✅	docker_dogstatsd_arm64	$${0}$$	$${37.06}$$ < $${37.94}$$	$${+0}$$	$${14.33}$$ < $${14.83}$$
✅	dogstatsd_deb_amd64	$${0}$$	$${29.98}$$ < $${30.61}$$	$${+0}$$	$${7.93}$$ < $${8.79}$$
✅	dogstatsd_deb_arm64	$${0}$$	$${28.14}$$ < $${29.11}$$	$${-0}$$	$${6.81}$$ < $${7.71}$$
✅	dogstatsd_rpm_amd64	$${0}$$	$${29.98}$$ < $${30.61}$$	$${+0}$$	$${7.94}$$ < $${8.8}$$
✅	dogstatsd_suse_amd64	$${0}$$	$${29.98}$$ < $${30.61}$$	$${+0}$$	$${7.94}$$ < $${8.8}$$
✅	iot_agent_deb_amd64	$${0}$$	$${42.92}$$ < $${43.29}$$	$${+0}$$	$${11.23}$$ < $${12.04}$$
✅	iot_agent_deb_arm64	$${0}$$	$${40.05}$$ < $${40.92}$$	$${+0}$$	$${9.61}$$ < $${10.45}$$
✅	iot_agent_deb_armhf	$${0}$$	$${40.63}$$ < $${41.03}$$	$${-0}$$	$${9.8}$$ < $${10.62}$$
✅	iot_agent_rpm_amd64	$${0}$$	$${42.92}$$ < $${43.29}$$	$${-0}$$	$${11.25}$$ < $${12.06}$$
✅	iot_agent_suse_amd64	$${0}$$	$${42.92}$$ < $${43.29}$$	$${-0}$$	$${11.25}$$ < $${12.06}$$

[cit-pr-commenter[bot]]

Regression Detector

Regression Detector Results

Metrics dashboard
Target profiles
Run ID: 00d65071-2358-478d-b1f7-1d08aaf94f21

Baseline: 69a72e41601e3d22eba91dae728760800a69c38d Comparison: 7653d00c252d1de8e28f797792a610cef28e6ffb Diff

Optimization Goals: ✅ No significant changes detected

Experiments ignored for regressions

Regressions in experiments with settings containing erratic: true are ignored.

perf	experiment	goal	Δ mean %	Δ mean % CI	trials	links
➖	docker_containers_cpu	% cpu utilization	+2.07	[-1.04, +5.19]	1	Logs

Fine details of change detection per experiment

perf	experiment	goal	Δ mean %	Δ mean % CI	trials	links
➖	docker_containers_cpu	% cpu utilization	+2.07	[-1.04, +5.19]	1	Logs
➖	ddot_metrics	memory utilization	+0.57	[+0.36, +0.79]	1	Logs
➖	file_tree	memory utilization	+0.47	[+0.42, +0.53]	1	Logs
➖	ddot_logs	memory utilization	+0.28	[+0.22, +0.34]	1	Logs
➖	docker_containers_memory	memory utilization	+0.27	[+0.20, +0.34]	1	Logs
➖	otlp_ingest_metrics	memory utilization	+0.16	[+0.01, +0.31]	1	Logs
➖	tcp_syslog_to_blackhole	ingress throughput	+0.10	[+0.02, +0.18]	1	Logs
➖	file_to_blackhole_0ms_latency	egress throughput	+0.10	[-0.32, +0.51]	1	Logs
➖	quality_gate_metrics_logs	memory utilization	+0.07	[-0.15, +0.28]	1	Logs bounds checks dashboard
➖	uds_dogstatsd_to_api_v3	ingress throughput	+0.02	[-0.11, +0.14]	1	Logs
➖	tcp_dd_logs_filter_exclude	ingress throughput	+0.00	[-0.07, +0.08]	1	Logs
➖	file_to_blackhole_100ms_latency	egress throughput	+0.00	[-0.05, +0.05]	1	Logs
➖	uds_dogstatsd_to_api	ingress throughput	-0.01	[-0.14, +0.12]	1	Logs
➖	file_to_blackhole_500ms_latency	egress throughput	-0.04	[-0.42, +0.33]	1	Logs
➖	uds_dogstatsd_20mb_12k_contexts_20_senders	memory utilization	-0.06	[-0.12, -0.00]	1	Logs
➖	file_to_blackhole_1000ms_latency	egress throughput	-0.08	[-0.48, +0.33]	1	Logs
➖	quality_gate_idle	memory utilization	-0.12	[-0.16, -0.07]	1	Logs bounds checks dashboard
➖	otlp_ingest_logs	memory utilization	-0.13	[-0.23, -0.03]	1	Logs
➖	quality_gate_idle_all_features	memory utilization	-0.31	[-0.35, -0.28]	1	Logs bounds checks dashboard
➖	ddot_metrics_sum_delta	memory utilization	-0.33	[-0.53, -0.13]	1	Logs
➖	ddot_metrics_sum_cumulativetodelta_exporter	memory utilization	-0.36	[-0.59, -0.14]	1	Logs
➖	quality_gate_logs	% cpu utilization	-0.44	[-1.94, +1.05]	1	Logs bounds checks dashboard
➖	ddot_metrics_sum_cumulative	memory utilization	-0.54	[-0.70, -0.38]	1	Logs

Bounds Checks: ✅ Passed

perf	experiment	bounds_check_name	replicates_passed	links
✅	docker_containers_cpu	simple_check_run	10/10
✅	docker_containers_memory	memory_usage	10/10
✅	docker_containers_memory	simple_check_run	10/10
✅	file_to_blackhole_0ms_latency	lost_bytes	10/10
✅	file_to_blackhole_0ms_latency	memory_usage	10/10
✅	file_to_blackhole_1000ms_latency	lost_bytes	10/10
✅	file_to_blackhole_1000ms_latency	memory_usage	10/10
✅	file_to_blackhole_100ms_latency	lost_bytes	10/10
✅	file_to_blackhole_100ms_latency	memory_usage	10/10
✅	file_to_blackhole_500ms_latency	lost_bytes	10/10
✅	file_to_blackhole_500ms_latency	memory_usage	10/10
✅	quality_gate_idle	intake_connections	10/10	bounds checks dashboard
✅	quality_gate_idle	memory_usage	10/10	bounds checks dashboard
✅	quality_gate_idle_all_features	intake_connections	10/10	bounds checks dashboard
✅	quality_gate_idle_all_features	memory_usage	10/10	bounds checks dashboard
✅	quality_gate_logs	intake_connections	10/10	bounds checks dashboard
✅	quality_gate_logs	lost_bytes	10/10	bounds checks dashboard
✅	quality_gate_logs	memory_usage	10/10	bounds checks dashboard
✅	quality_gate_metrics_logs	cpu_usage	10/10	bounds checks dashboard
✅	quality_gate_metrics_logs	intake_connections	10/10	bounds checks dashboard
✅	quality_gate_metrics_logs	lost_bytes	10/10	bounds checks dashboard
✅	quality_gate_metrics_logs	memory_usage	10/10	bounds checks dashboard

Explanation

Confidence level: 90.00% Effect size tolerance: |Δ mean %| ≥ 5.00%

Performance changes are noted in the perf column of each table:

• ✅ = significantly better comparison variant performance
• ❌ = significantly worse comparison variant performance
• ➖ = no significant change in performance

A regression test is an A/B test of target performance in a repeatable rig, where "performance" is measured as "comparison variant minus baseline variant" for an optimization goal (e.g., ingress throughput). Due to intrinsic variability in measuring that goal, we can only estimate its mean value for each experiment; we report uncertainty in that value as a 90.00% confidence interval denoted "Δ mean % CI".

For each experiment, we decide whether a change in performance is a "regression" -- a change worth investigating further -- if all of the following criteria are true:

1. Its estimated |Δ mean %| ≥ 5.00%, indicating the change is big enough to merit a closer look.

2. Its 90.00% confidence interval "Δ mean % CI" does not contain zero, indicating that if our statistical model is accurate, there is at least a 90.00% chance there is a difference in performance between baseline and comparison variants.

3. Its configuration does not mark it "erratic".

CI Pass/Fail Decision

✅ Passed. All Quality Gates passed.

• quality_gate_idle, bounds check memory_usage: 10/10 replicas passed. Gate passed.
• quality_gate_idle, bounds check intake_connections: 10/10 replicas passed. Gate passed.
• quality_gate_logs, bounds check memory_usage: 10/10 replicas passed. Gate passed.
• quality_gate_logs, bounds check lost_bytes: 10/10 replicas passed. Gate passed.
• quality_gate_logs, bounds check intake_connections: 10/10 replicas passed. Gate passed.
• quality_gate_metrics_logs, bounds check lost_bytes: 10/10 replicas passed. Gate passed.
• quality_gate_metrics_logs, bounds check memory_usage: 10/10 replicas passed. Gate passed.
• quality_gate_metrics_logs, bounds check intake_connections: 10/10 replicas passed. Gate passed.
• quality_gate_metrics_logs, bounds check cpu_usage: 10/10 replicas passed. Gate passed.
• quality_gate_idle_all_features, bounds check memory_usage: 10/10 replicas passed. Gate passed.
• quality_gate_idle_all_features, bounds check intake_connections: 10/10 replicas passed. Gate passed.

[gjulianm]

/trigger-ci

[dd-devflow-routing-codex[bot]]

View all feedbacks in Devflow UI.

2025-12-01 10:15:28 UTC :information_source: Start processing command /trigger-ci If you need support, contact us on Slack #devflow!

---

2025-12-01 10:15:45 UTC :information_source: Branch synced on Gitlab

Branch guillermo.julian/gpu-ci-tests was out-of-sync on Gitlab - it has been resynced

If you need support, contact us on Slack #devflow!

---

2025-12-01 10:17:26 UTC :rotating_light: Devflow

POST https://gitlab.ddbuild.io/api/v4/projects/DataDog/datadog-agent/pipeline: 400 {message: {base: [jobs:tests_gpu config contains unknown keys: vars]}}

Details

child workflow execution error (type: devtools.Devtools_TriggerCiFromPr, workflowID: ef11e0ec-3b2e-4473-9982-53de1283db31_36, runID: a0513b8e-2583-41ff-9834-75de66b44a62, initiatedEventID: 36, startedEventID: 37): child workflow execution error (type: devtools.Devtools_TriggerCi, workflowID: a0513b8e-2583-41ff-9834-75de66b44a62_14, runID: 4cd4a309-6a45-4496-a1c5-f1a567d07c78, initiatedEventID: 14, startedEventID: 15): activity error (type: gitlab.GitlabService_CreatePipeline, scheduledEventID: 55, startedEventID: 56, identity: 1@gitlab-worker-8d9795d5-57fks@): POST https://gitlab.ddbuild.io/api/v4/projects/DataDog/datadog-agent/pipeline: 400 {message: {base: [jobs:tests_gpu config contains unknown keys: vars]}} (type: ErrorResponse, retryable: true)

If you need support, contact us on Slack #devflow with those details!

[chouetz]

@codex review

[chatgpt-codex-connector[bot]]

To use Codex here, create a Codex account and connect to github.

[gjulianm]

/merge

[dd-devflow-routing-codex[bot]]

View all feedbacks in Devflow UI.

2025-12-18 14:07:21 UTC :information_source: Start processing command /merge

---

2025-12-18 14:07:29 UTC :information_source: MergeQueue: waiting for PR to be ready

This pull request is not mergeable according to GitHub. Common reasons include pending required checks, missing approvals, or merge conflicts — but it could also be blocked by other repository rules or settings. It will be added to the queue as soon as checks pass and/or get approvals. Note: if you pushed new commits since the last approval, you may need additional approval. You can remove it from the waiting list with /remove command.

---

2025-12-18 15:03:16 UTC :information_source: MergeQueue: merge request added to the queue

The expected merge time in main is approximately 1h (p90).

---

2025-12-18 15:44:06 UTC :information_source: MergeQueue: This merge request was merged