[EBPF] gpu: add validation task against live metrics

[gjulianm]

What does this PR do?

Adds task gpu.validate-metrics to check the specification against metrics sent to Datadog.

Most of the code is placed in tasks/libs/gpu for separation, additionally it guards against importing packages that are not present by default in dda installation.

Motivation

Help validation by ensuring the behavior of the agent in production matches what is in the specification.

Describe how you validated your changes

Tested manually, only python code for auxiliary tasks, no agent code changes.

Additional Notes

Some of the metric specifications are updated to match what's being actually emitted.

There were also some changes in pyproject.toml to ensure typing is enforced in the new code, and to avoid a vulture error with pydantic schemas.

[agent-platform-auto-pr]

Files inventory check summary

File checks results against ancestor 7d01f6de:

Results for datadog-agent_7.79.0~devel.git.187.1cde0e8.pipeline.104495341-1_amd64.deb:

No change detected

[agent-platform-auto-pr]

Static quality checks

✅ Please find below the results from static quality gates
Comparison made with ancestor 7d01f6d
📊 Static Quality Gates Dashboard
🔗 SQG Job

30 successful checks with minimal change (< 2 KiB)

	Quality gate	Current Size
✅	agent_deb_amd64	752.537 MiB
✅	agent_deb_amd64_fips	709.559 MiB
✅	agent_heroku_amd64	313.293 MiB
✅	agent_rpm_amd64	752.521 MiB
✅	agent_rpm_amd64_fips	709.542 MiB
✅	agent_rpm_arm64	730.963 MiB
✅	agent_rpm_arm64_fips	690.994 MiB
✅	agent_suse_amd64	752.521 MiB
✅	agent_suse_amd64_fips	709.542 MiB
✅	agent_suse_arm64	730.963 MiB
✅	agent_suse_arm64_fips	690.994 MiB
✅	docker_agent_amd64	812.839 MiB
✅	docker_agent_arm64	816.052 MiB
✅	docker_agent_jmx_amd64	1003.755 MiB
✅	docker_agent_jmx_arm64	995.746 MiB
✅	docker_cluster_agent_amd64	203.938 MiB
✅	docker_cluster_agent_arm64	218.420 MiB
✅	docker_cws_instrumentation_amd64	7.142 MiB
✅	docker_cws_instrumentation_arm64	6.689 MiB
✅	docker_dogstatsd_amd64	39.238 MiB
✅	docker_dogstatsd_arm64	37.445 MiB
✅	dogstatsd_deb_amd64	29.881 MiB
✅	dogstatsd_deb_arm64	28.030 MiB
✅	dogstatsd_rpm_amd64	29.881 MiB
✅	dogstatsd_suse_amd64	29.881 MiB
✅	iot_agent_deb_amd64	43.270 MiB
✅	iot_agent_deb_arm64	40.320 MiB
✅	iot_agent_deb_armhf	41.064 MiB
✅	iot_agent_rpm_amd64	43.270 MiB
✅	iot_agent_suse_amd64	43.270 MiB

On-wire sizes (compressed)

	Quality gate	Change	Size (prev → curr → max)
✅	agent_deb_amd64	-10.4 KiB (0.01% reduction)	174.711 → 174.700 → 178.360
✅	agent_deb_amd64_fips	-20.69 KiB (0.01% reduction)	165.311 → 165.291 → 172.790
✅	agent_heroku_amd64	neutral	75.002 MiB → 79.970
✅	agent_rpm_amd64	-6.66 KiB (0.00% reduction)	177.540 → 177.533 → 181.830
✅	agent_rpm_amd64_fips	neutral	167.596 MiB → 173.370
✅	agent_rpm_arm64	-39.6 KiB (0.02% reduction)	159.523 → 159.485 → 163.060
✅	agent_rpm_arm64_fips	-14.45 KiB (0.01% reduction)	151.334 → 151.320 → 156.170
✅	agent_suse_amd64	-6.66 KiB (0.00% reduction)	177.540 → 177.533 → 181.830
✅	agent_suse_amd64_fips	neutral	167.596 MiB → 173.370
✅	agent_suse_arm64	-39.6 KiB (0.02% reduction)	159.523 → 159.485 → 163.060
✅	agent_suse_arm64_fips	-14.45 KiB (0.01% reduction)	151.334 → 151.320 → 156.170
✅	docker_agent_amd64	neutral	268.093 MiB → 272.480
✅	docker_agent_arm64	neutral	255.304 MiB → 261.060
✅	docker_agent_jmx_amd64	+3.02 KiB (0.00% increase)	336.747 → 336.750 → 341.100
✅	docker_agent_jmx_arm64	neutral	319.932 MiB → 325.620
✅	docker_cluster_agent_amd64	neutral	71.367 MiB → 72.920
✅	docker_cluster_agent_arm64	neutral	67.015 MiB → 68.220
✅	docker_cws_instrumentation_amd64	neutral	2.999 MiB → 3.330
✅	docker_cws_instrumentation_arm64	neutral	2.729 MiB → 3.090
✅	docker_dogstatsd_amd64	neutral	15.173 MiB → 15.820
✅	docker_dogstatsd_arm64	neutral	14.487 MiB → 14.830
✅	dogstatsd_deb_amd64	neutral	7.893 MiB → 8.790
✅	dogstatsd_deb_arm64	neutral	6.777 MiB → 7.710
✅	dogstatsd_rpm_amd64	neutral	7.903 MiB → 8.800
✅	dogstatsd_suse_amd64	neutral	7.903 MiB → 8.800
✅	iot_agent_deb_amd64	neutral	11.398 MiB → 12.040
✅	iot_agent_deb_arm64	neutral	9.702 MiB → 10.450
✅	iot_agent_deb_armhf	neutral	9.939 MiB → 10.620
✅	iot_agent_rpm_amd64	neutral	11.417 MiB → 12.060
✅	iot_agent_suse_amd64	neutral	11.417 MiB → 12.060

[cit-pr-commenter-54b7da]

Regression Detector

Regression Detector Results

Metrics dashboard
Target profiles
Run ID: f830af84-0d8b-4c4f-a0f7-083e5e576c7e

Baseline: 7d01f6d
Comparison: 1cde0e8
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	+5.56	[+2.50, +8.62]	1	Logs

Fine details of change detection per experiment

perf	experiment	goal	Δ mean %	Δ mean % CI	trials	links
❌	docker_containers_cpu	% cpu utilization	+5.56	[+2.50, +8.62]	1	Logs
➖	quality_gate_logs	% cpu utilization	+0.84	[-0.81, +2.48]	1	Logs bounds checks dashboard
➖	ddot_metrics	memory utilization	+0.29	[+0.12, +0.47]	1	Logs
➖	ddot_metrics_sum_delta	memory utilization	+0.26	[+0.09, +0.43]	1	Logs
➖	docker_containers_memory	memory utilization	+0.15	[+0.07, +0.23]	1	Logs
➖	quality_gate_metrics_logs	memory utilization	+0.08	[-0.15, +0.32]	1	Logs bounds checks dashboard
➖	quality_gate_idle_all_features	memory utilization	+0.08	[+0.04, +0.11]	1	Logs bounds checks dashboard
➖	uds_dogstatsd_20mb_12k_contexts_20_senders	memory utilization	+0.04	[-0.02, +0.10]	1	Logs
➖	file_to_blackhole_1000ms_latency	egress throughput	+0.02	[-0.40, +0.45]	1	Logs
➖	uds_dogstatsd_to_api	ingress throughput	+0.01	[-0.19, +0.21]	1	Logs
➖	tcp_dd_logs_filter_exclude	ingress throughput	+0.00	[-0.12, +0.12]	1	Logs
➖	ddot_logs	memory utilization	-0.01	[-0.07, +0.05]	1	Logs
➖	file_to_blackhole_500ms_latency	egress throughput	-0.02	[-0.41, +0.38]	1	Logs
➖	uds_dogstatsd_to_api_v3	ingress throughput	-0.02	[-0.22, +0.18]	1	Logs
➖	file_to_blackhole_100ms_latency	egress throughput	-0.06	[-0.13, +0.02]	1	Logs
➖	ddot_metrics_sum_cumulativetodelta_exporter	memory utilization	-0.07	[-0.29, +0.16]	1	Logs
➖	file_to_blackhole_0ms_latency	egress throughput	-0.08	[-0.60, +0.44]	1	Logs
➖	file_tree	memory utilization	-0.11	[-0.17, -0.05]	1	Logs
➖	ddot_metrics_sum_cumulative	memory utilization	-0.11	[-0.26, +0.03]	1	Logs
➖	tcp_syslog_to_blackhole	ingress throughput	-0.14	[-0.29, +0.00]	1	Logs
➖	otlp_ingest_metrics	memory utilization	-0.18	[-0.34, -0.01]	1	Logs
➖	otlp_ingest_logs	memory utilization	-0.32	[-0.42, -0.22]	1	Logs
➖	quality_gate_idle	memory utilization	-0.32	[-0.37, -0.27]	1	Logs bounds checks dashboard

Bounds Checks: ✅ Passed

perf	experiment	bounds_check_name	replicates_passed	observed_value	links
✅	docker_containers_cpu	simple_check_run	10/10	672 ≥ 26
✅	docker_containers_memory	memory_usage	10/10	271.17MiB ≤ 370MiB
✅	docker_containers_memory	simple_check_run	10/10	696 ≥ 26
✅	file_to_blackhole_0ms_latency	memory_usage	10/10	0.19GiB ≤ 1.20GiB
✅	file_to_blackhole_0ms_latency	missed_bytes	10/10	0B = 0B
✅	file_to_blackhole_1000ms_latency	memory_usage	10/10	0.23GiB ≤ 1.20GiB
✅	file_to_blackhole_1000ms_latency	missed_bytes	10/10	0B = 0B
✅	file_to_blackhole_100ms_latency	memory_usage	10/10	0.19GiB ≤ 1.20GiB
✅	file_to_blackhole_100ms_latency	missed_bytes	10/10	0B = 0B
✅	file_to_blackhole_500ms_latency	memory_usage	10/10	0.21GiB ≤ 1.20GiB
✅	file_to_blackhole_500ms_latency	missed_bytes	10/10	0B = 0B
✅	quality_gate_idle	intake_connections	10/10	3 = 3	bounds checks dashboard
✅	quality_gate_idle	memory_usage	10/10	173.14MiB ≤ 175MiB	bounds checks dashboard
✅	quality_gate_idle_all_features	intake_connections	10/10	3 = 3	bounds checks dashboard
✅	quality_gate_idle_all_features	memory_usage	10/10	500.08MiB ≤ 550MiB	bounds checks dashboard
✅	quality_gate_logs	intake_connections	10/10	4 ≤ 6	bounds checks dashboard
✅	quality_gate_logs	memory_usage	10/10	207.16MiB ≤ 220MiB	bounds checks dashboard
✅	quality_gate_logs	missed_bytes	10/10	0B = 0B	bounds checks dashboard
✅	quality_gate_metrics_logs	cpu_usage	10/10	365.85 ≤ 2000	bounds checks dashboard
✅	quality_gate_metrics_logs	intake_connections	10/10	4 ≤ 6	bounds checks dashboard
✅	quality_gate_metrics_logs	memory_usage	10/10	414.12MiB ≤ 475MiB	bounds checks dashboard
✅	quality_gate_metrics_logs	missed_bytes	10/10	0B = 0B	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_metrics_logs, bounds check memory_usage: 10/10 replicas passed. Gate passed.
• quality_gate_metrics_logs, bounds check cpu_usage: 10/10 replicas passed. Gate passed.
• quality_gate_metrics_logs, bounds check missed_bytes: 10/10 replicas passed. Gate passed.
• quality_gate_metrics_logs, bounds check intake_connections: 10/10 replicas passed. Gate 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_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.
• quality_gate_logs, bounds check memory_usage: 10/10 replicas passed. Gate passed.
• quality_gate_logs, bounds check missed_bytes: 10/10 replicas passed. Gate passed.
• quality_gate_logs, bounds check intake_connections: 10/10 replicas passed. Gate passed.

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: f6675d685f

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[chatgpt-codex-connector]

Classify host_vgpu as vgpu device mode

normalize_device_mode only maps virtualization_mode == "vgpu" to vGPU, so values like host_vgpu are treated as physical. The NVML collector can emit host_vgpu for vGPU devices, so this misclassifies live configs and validates them against the physical support matrix, producing false failures and hiding real vGPU gaps.

Useful? React with 👍 / 👎.

[gjulianm]

host_vgpu is not supported yet and we are not clear on the type of metrics that will appear in that case. For now, not treating it as a separate case.

[gjulianm]

/merge

[gh-worker-devflow-routing-ef8351]

View all feedbacks in Devflow UI.

2026-03-26 15:56:39 UTC ℹ️ Start processing command /merge

---

2026-03-26 15:56:44 UTC ℹ️ MergeQueue: pull request added to the queue

The expected merge time in main is approximately 2h (p90).

---

2026-03-26 16:39:35 UTC ℹ️ MergeQueue: This merge request was merged