[EBPF-641] gpu: configure cgroup permissions for agent #33588

gjulianm · 2025-01-30T16:16:31Z

What does this PR do?

Updates the cgroup permission configuration in system-probe to include the agent process, so that it will be given access to the NVIDIA devices.

Motivation

Avoid errors due to missing permissions for opening NVIDIA devices. More details in RFC

https://datadoghq.atlassian.net/browse/EBPF-642
https://datadoghq.atlassian.net/browse/EBPF-641

Describe how you validated your changes

Validated locally and in staging clusters that the permissions are correctly set.

Possible Drawbacks / Trade-offs

Additional Notes

agent-platform-auto-pr · 2025-01-30T16:53:00Z

Uncompressed package size comparison

Comparison with ancestor 9a49faf510b46f65d4d109e15afb9c294d7ff890

Diff per package

package	diff	status	size	ancestor	threshold
datadog-agent-x86_64-rpm	0.01MB	⚠️	886.81MB	886.81MB	0.50MB
datadog-agent-x86_64-suse	0.01MB	⚠️	886.81MB	886.81MB	0.50MB
datadog-agent-amd64-deb	0.01MB	⚠️	877.08MB	877.07MB	0.50MB
datadog-iot-agent-amd64-deb	0.00MB	✅	93.87MB	93.87MB	0.50MB
datadog-iot-agent-x86_64-rpm	0.00MB	✅	93.94MB	93.94MB	0.50MB
datadog-iot-agent-x86_64-suse	0.00MB	✅	93.94MB	93.94MB	0.50MB
datadog-iot-agent-aarch64-rpm	0.00MB	✅	89.99MB	89.99MB	0.50MB
datadog-agent-arm64-deb	0.00MB	✅	864.87MB	864.87MB	0.50MB
datadog-dogstatsd-amd64-deb	0.00MB	✅	59.02MB	59.02MB	0.50MB
datadog-dogstatsd-x86_64-rpm	0.00MB	✅	59.10MB	59.10MB	0.50MB
datadog-dogstatsd-x86_64-suse	0.00MB	✅	59.10MB	59.10MB	0.50MB
datadog-dogstatsd-arm64-deb	0.00MB	✅	56.51MB	56.51MB	0.50MB
datadog-heroku-agent-amd64-deb	0.00MB	✅	456.50MB	456.50MB	0.50MB
datadog-iot-agent-arm64-deb	0.00MB	✅	89.92MB	89.92MB	0.50MB
datadog-agent-aarch64-rpm	-0.00MB	✅	874.59MB	874.59MB	0.50MB

Decision

⚠️ Warning

agent-platform-auto-pr · 2025-01-30T16:55:44Z

Test changes on VM

Use this command from test-infra-definitions to manually test this PR changes on a VM:

inv aws.create-vm --pipeline-id=54724393 --os-family=ubuntu

Note: This applies to commit c13dc5b

cit-pr-commenter · 2025-01-30T17:23:21Z

Regression Detector

Regression Detector Results

Metrics dashboard
Target profiles
Run ID: 75924c5e-c498-469f-be0b-cfff2e991d60

Baseline: 9a49faf
Comparison: c13dc5b
Diff

Optimization Goals: ✅ No significant changes detected

Fine details of change detection per experiment

perf	experiment	goal	Δ mean %	Δ mean % CI	trials	links
➖	file_to_blackhole_1000ms_latency	egress throughput	+0.50	[-0.28, +1.27]	1	Logs
➖	quality_gate_idle_all_features	memory utilization	+0.45	[+0.37, +0.53]	1	Logs bounds checks dashboard
➖	file_tree	memory utilization	+0.43	[+0.37, +0.49]	1	Logs
➖	quality_gate_idle	memory utilization	+0.10	[+0.07, +0.14]	1	Logs bounds checks dashboard
➖	file_to_blackhole_0ms_latency	egress throughput	+0.02	[-0.85, +0.89]	1	Logs
➖	file_to_blackhole_0ms_latency_http1	egress throughput	+0.02	[-0.88, +0.91]	1	Logs
➖	file_to_blackhole_0ms_latency_http2	egress throughput	+0.01	[-0.87, +0.89]	1	Logs
➖	file_to_blackhole_300ms_latency	egress throughput	+0.01	[-0.63, +0.65]	1	Logs
➖	tcp_dd_logs_filter_exclude	ingress throughput	-0.01	[-0.04, +0.03]	1	Logs
➖	file_to_blackhole_100ms_latency	egress throughput	-0.01	[-0.74, +0.73]	1	Logs
➖	uds_dogstatsd_to_api	ingress throughput	-0.01	[-0.29, +0.28]	1	Logs
➖	file_to_blackhole_1000ms_latency_linear_load	egress throughput	-0.02	[-0.48, +0.45]	1	Logs
➖	file_to_blackhole_500ms_latency	egress throughput	-0.11	[-0.89, +0.68]	1	Logs
➖	tcp_syslog_to_blackhole	ingress throughput	-1.13	[-1.31, -0.96]	1	Logs
➖	uds_dogstatsd_to_api_cpu	% cpu utilization	-1.31	[-2.20, -0.42]	1	Logs
➖	quality_gate_logs	% cpu utilization	-2.30	[-5.33, +0.72]	1	Logs

Bounds Checks: ✅ Passed

perf	experiment	bounds_check_name	replicates_passed	links
✅	file_to_blackhole_0ms_latency	lost_bytes	10/10
✅	file_to_blackhole_0ms_latency	memory_usage	10/10
✅	file_to_blackhole_0ms_latency_http1	lost_bytes	10/10
✅	file_to_blackhole_0ms_latency_http1	memory_usage	10/10
✅	file_to_blackhole_0ms_latency_http2	lost_bytes	10/10
✅	file_to_blackhole_0ms_latency_http2	memory_usage	10/10
✅	file_to_blackhole_1000ms_latency	memory_usage	10/10
✅	file_to_blackhole_1000ms_latency_linear_load	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_300ms_latency	lost_bytes	10/10
✅	file_to_blackhole_300ms_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
✅	quality_gate_logs	lost_bytes	10/10
✅	quality_gate_logs	memory_usage	10/10

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:

Its estimated |Δ mean %| ≥ 5.00%, indicating the change is big enough to merit a closer look.
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.
Its configuration does not mark it "erratic".

CI Pass/Fail Decision

✅ Passed. All Quality Gates passed.

quality_gate_idle, 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_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 lost_bytes: 10/10 replicas passed. Gate passed.
quality_gate_logs, bounds check intake_connections: 10/10 replicas passed. Gate passed.
quality_gate_logs, bounds check memory_usage: 10/10 replicas passed. Gate passed.

val06

reviewed

val06 · 2025-02-03T12:04:45Z

cmd/system-probe/modules/gpu.go

+	root := hostRoot()
+
+	sysprobePID := uint32(os.Getpid())
+	log.Infof("Configuring cgroup permissions for system-probe process with PID %d", sysprobePID)


nit: maybe change this to trace level?

I'd rather leave it as info for now, as we'll probably want to check if the PIDs are correct just in case.

cmd/system-probe/modules/gpu.go

comp/core/workloadmeta/collectors/internal/nvml/nvml.go

val06 · 2025-02-03T12:10:23Z

cmd/system-probe/modules/gpu.go

+	agentPID, err := getAgentPID(procRoot)
+	if err != nil {
+		log.Warnf("Failed to get agent PID: %v. Cannot patch cgroup permissions, gpu-monitoring module might not work properly", err)
+		return


why do we return only in this case?

Because with no PID we cannot set the permissions, in the other cases we do want to continue (e.g., failing to set system-probe permissions doesn't mean we cannot try with the agent anyways).

gjulianm · 2025-02-03T15:07:32Z

/merge

gjulianm · 2025-02-03T15:07:48Z

/merge -c

dd-devflow · 2025-02-03T15:07:50Z

Devflow running: `/merge`

View all feedbacks in Devflow UI.

2025-02-03 15:07:50 UTC ℹ️ MergeQueue: waiting for PR to be ready

This merge request is not mergeable yet, because of pending checks/missing approvals. 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.

2025-02-03 15:07:57 UTC ⚠️ MergeQueue: This merge request was unqueued

[email protected] unqueued this merge request

gjulianm · 2025-02-03T18:17:37Z

/merge

dd-devflow · 2025-02-03T18:17:49Z

Devflow running: `/merge`

View all feedbacks in Devflow UI.

2025-02-03 18:17:48 UTC ℹ️ MergeQueue: pull request added to the queue

The median merge time in main is 27m.

2025-02-03 18:50:02 UTC ℹ️ MergeQueue: This merge request was merged

gjulianm self-assigned this Jan 30, 2025

github-actions bot added component/system-probe short review PR is simple enough to be reviewed quickly team/ebpf-platform labels Jan 30, 2025

gjulianm added 2 commits January 31, 2025 12:37

Patch agent permissions too

8702add

Retry NVML usage in agent components

e6fdd21

gjulianm force-pushed the guillermo.julian/fix-agent-permissions branch from 3f4f3b2 to e6fdd21 Compare January 31, 2025 11:38

gjulianm added changelog/no-changelog qa/done QA done before merge and regressions are covered by tests labels Jan 31, 2025

gjulianm marked this pull request as ready for review February 3, 2025 12:01

gjulianm requested a review from a team as a code owner February 3, 2025 12:01

gjulianm added the ask-review Ask required teams to review this PR label Feb 3, 2025

val06 reviewed Feb 3, 2025

View reviewed changes

PR comments

c4c65d3

val06 approved these changes Feb 3, 2025

View reviewed changes

Merge branch 'main' into guillermo.julian/fix-agent-permissions

c13dc5b

github-actions bot added medium review PR review might take time and removed short review PR is simple enough to be reviewed quickly labels Feb 3, 2025

gjulianm changed the title ~~[EBPF] gpu: configure cgroup permissions for agent~~ [EBPF-641] gpu: configure cgroup permissions for agent Feb 3, 2025

dd-mergequeue bot merged commit 22c847f into main Feb 3, 2025
328 checks passed

dd-mergequeue bot deleted the guillermo.julian/fix-agent-permissions branch February 3, 2025 18:50

github-actions bot added this to the 7.64.0 milestone Feb 3, 2025

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

[EBPF-641] gpu: configure cgroup permissions for agent #33588

[EBPF-641] gpu: configure cgroup permissions for agent #33588

gjulianm commented Jan 30, 2025 •

edited

Loading

agent-platform-auto-pr bot commented Jan 30, 2025 •

edited

Loading

agent-platform-auto-pr bot commented Jan 30, 2025 •

edited

Loading

cit-pr-commenter bot commented Jan 30, 2025 •

edited

Loading

Fine details of change detection per experiment

Bounds Checks: ✅ Passed

Explanation

val06 left a comment

val06 Feb 3, 2025

gjulianm Feb 3, 2025

val06 Feb 3, 2025

gjulianm Feb 3, 2025

gjulianm commented Feb 3, 2025

gjulianm commented Feb 3, 2025

dd-devflow bot commented Feb 3, 2025 •

edited

Loading

gjulianm commented Feb 3, 2025

dd-devflow bot commented Feb 3, 2025 •

edited

Loading

[EBPF-641] gpu: configure cgroup permissions for agent #33588

[EBPF-641] gpu: configure cgroup permissions for agent #33588

Conversation

gjulianm commented Jan 30, 2025 • edited Loading

What does this PR do?

Motivation

Describe how you validated your changes

Possible Drawbacks / Trade-offs

Additional Notes

agent-platform-auto-pr bot commented Jan 30, 2025 • edited Loading

Uncompressed package size comparison

Decision

agent-platform-auto-pr bot commented Jan 30, 2025 • edited Loading

Test changes on VM

cit-pr-commenter bot commented Jan 30, 2025 • edited Loading

Regression Detector

Regression Detector Results

Optimization Goals: ✅ No significant changes detected

Fine details of change detection per experiment

Bounds Checks: ✅ Passed

Explanation

CI Pass/Fail Decision

val06 left a comment

Choose a reason for hiding this comment

val06 Feb 3, 2025

Choose a reason for hiding this comment

gjulianm Feb 3, 2025

Choose a reason for hiding this comment

val06 Feb 3, 2025

Choose a reason for hiding this comment

gjulianm Feb 3, 2025

Choose a reason for hiding this comment

gjulianm commented Feb 3, 2025

gjulianm commented Feb 3, 2025

dd-devflow bot commented Feb 3, 2025 • edited Loading

Devflow running: /merge

gjulianm commented Feb 3, 2025

dd-devflow bot commented Feb 3, 2025 • edited Loading

Devflow running: /merge

gjulianm commented Jan 30, 2025 •

edited

Loading

agent-platform-auto-pr bot commented Jan 30, 2025 •

edited

Loading

agent-platform-auto-pr bot commented Jan 30, 2025 •

edited

Loading

cit-pr-commenter bot commented Jan 30, 2025 •

edited

Loading

dd-devflow bot commented Feb 3, 2025 •

edited

Loading

Devflow running: `/merge`

dd-devflow bot commented Feb 3, 2025 •

edited

Loading

Devflow running: `/merge`