feat(network): attribute proxy sockets to the pod owning their netns

[matthewleese]

What does this PR do?

Adds network_config.enable_netns_container_attribution to system-probe, off by default. When it's on, a connection whose socket was created by an allowlisted proxy inside another pod's network namespace is attributed to the container that owns that namespace, not to the proxy.

• pkg/network/tracer/netns_containers.go: a map of network namespace inode to container, fed by workloadmeta container events. Each container's namespace is read once from /proc/<pid>/ns/net when it starts, and removed when it's deleted, because inodes can be reused. The host's namespace is never mapped.
• addProcessInfo: after the process cache sets the container, reattribute when the owner's image is on network_config.netns_container_attribution_owner_images (default ztunnel), the namespace holds exactly one container, and that container isn't the owner. Multi-container pods keep today's behavior. Reattributed connections drop the owner process's service, env, and version tags. This runs before aggregation, because the connection key includes the container.
• The network module enables it after the tracer starts; Windows and macOS are no-ops.

Remote ends need no change: every remote-container resolver copies the peer socket's container, so once the local end is right the remote end follows.

Motivation

In Istio ambient mode, ztunnel calls setns into each enrolled pod's network namespace before bind and connect. Its sockets live in the app pod's namespace, but they're owned by ztunnel's process, so CNM attributes all of that traffic to ztunnel. A customer running ambient sees ztunnel as their largest client and can't see which services talk to each other.

RFC: [RFC] CNM Ambient ztunnel Mitigation for NYT

Describe how you validated your changes

• Unit tests for the resolver: reattribution, own namespace, non-owners, multi-container pods, host and unknown namespaces, delete then inode reuse, stopped and unreadable containers, and concurrent updates under -race.
• TestNetNSContainerAttribution (TracerSuite): the test process enters a new namespace with netns.WithNS, dials a loopback server there, and asserts the connection is attributed to that namespace's container with process tags cleared, while a control connection in its own namespace keeps the owner's container. Compiled locally; not yet run on a kernel.
• Built for Linux (bpf) and macOS locally. Not yet validated in an ambient cluster.

Additional Notes

• Draft: still needs the kernel test run in CI and an ambient lab run.
• Known limitation: containerForPid can map the proxy's PID to one of the reattributed containers. The backend only falls back to it when a connection has no container ID, which for a proxy socket happens only on a process-cache miss.
• A follow-up will add a payload field marking reattributed connections, so the backend can label and hide the hops inside a pod.

Note: This PR was created by Claude.

🤖 Generated with Claude Code

[github-actions]

AI review by Codex (OpenAI) - workflow run

The patch is incorrect: the new resolver is missing from the Bazel target, breaking Linux/BPF builds. Review was limited to static inspection; tests were not run.

[github-actions]

AI review by Codex (OpenAI) - workflow run

The patch is incorrect: payload enrichment can restore the proxy’s tags, and rewriting the source container also corrupts the PID-to-container fallback mapping. The kernel test does not cover either downstream behavior.

[datadog-datadog-prod-us1-2]

🎯 Code Coverage (details)
• Patch Coverage: 81.40%
• Overall Coverage: 58.59% (+0.00%)

_{This comment will be updated automatically if new data arrives.
🔗 Commit SHA: 41549ad | Docs | Give us feedback!}

[dd-octo-sts]

Files inventory check summary

File checks results against ancestor 1e25e036:

Results for datadog-agent_7.86.0~devel.git.306.41549ad.pipeline.143647082-1_amd64.deb:

No change detected

Results for datadog-iot-agent_7.86.0~devel.git.306.41549ad.pipeline.143647082-1_amd64.deb:

No change detected

[dd-octo-sts]

Static quality checks

✅ Please find below the results from static quality gates
Comparison made with ancestor 1e25e03
📊 Static Quality Gates Dashboard
🔗 SQG Job

Successful checks

Info

	Quality gate	Change	Size (prev → curr → max)
✅	agent_deb_amd64	+12.33 KiB (0.00% increase, -0.23% of buffer)	755.264 → 755.276 → 760.520
✅	agent_deb_amd64_fips	+28.0 KiB (0.00% increase, -0.52% of buffer)	688.274 → 688.301 → 693.570
✅	agent_msi	+8.0 KiB (0.00% increase, -2.17% of buffer)	653.830 → 653.838 → 654.190
✅	agent_rpm_amd64	+12.33 KiB (0.00% increase, -0.23% of buffer)	755.247 → 755.259 → 760.500
✅	agent_rpm_amd64_fips	+28.0 KiB (0.00% increase, -0.52% of buffer)	688.258 → 688.285 → 693.550
✅	agent_rpm_arm64	+19.99 KiB (0.00% increase, -0.37% of buffer)	725.785 → 725.804 → 731.100
✅	agent_rpm_arm64_fips	+20.0 KiB (0.00% increase, -0.36% of buffer)	664.367 → 664.387 → 669.740
✅	agent_suse_amd64	+12.33 KiB (0.00% increase, -0.23% of buffer)	755.247 → 755.259 → 760.500
✅	agent_suse_amd64_fips	+28.0 KiB (0.00% increase, -0.52% of buffer)	688.258 → 688.285 → 693.550
✅	agent_suse_arm64	+19.99 KiB (0.00% increase, -0.37% of buffer)	725.785 → 725.804 → 731.100
✅	agent_suse_arm64_fips	+20.0 KiB (0.00% increase, -0.36% of buffer)	664.367 → 664.387 → 669.740
✅	docker_agent_amd64	+12.34 KiB (0.00% increase, -0.23% of buffer)	811.427 → 811.439 → 816.720
✅	docker_agent_arm64	+19.99 KiB (0.00% increase, -0.36% of buffer)	806.591 → 806.610 → 811.970
✅	docker_agent_jmx_amd64	+12.33 KiB (0.00% increase, -0.23% of buffer)	1002.296 → 1002.308 → 1007.590
✅	docker_agent_jmx_arm64	+19.99 KiB (0.00% increase, -0.36% of buffer)	986.240 → 986.260 → 991.620
✅	dogstatsd_deb_arm64	+4.0 KiB (0.01% increase, -0.42% of buffer)	28.278 → 28.282 → 29.200
✅	iot_agent_deb_amd64	+4.0 KiB (0.01% increase, -2.18% of buffer)	47.371 → 47.375 → 47.550
✅	iot_agent_deb_arm64	+4.0 KiB (0.01% increase, -1.02% of buffer)	43.837 → 43.841 → 44.220
✅	iot_agent_rpm_amd64	+4.0 KiB (0.01% increase, -2.19% of buffer)	47.371 → 47.375 → 47.550
✅	iot_agent_suse_amd64	+4.0 KiB (0.01% increase, -2.17% of buffer)	47.370 → 47.374 → 47.550

13 successful checks with minimal change (< 2 KiB)

	Quality gate	Current Size
✅	agent_heroku_amd64	319.144 MiB
✅	docker_cws_instrumentation_amd64	7.443 MiB
✅	docker_cws_instrumentation_arm64	6.877 MiB
✅	docker_dogstatsd_amd64	39.551 MiB
✅	docker_dogstatsd_arm64	37.629 MiB
✅	docker_host_profiler_amd64	308.721 MiB
✅	docker_host_profiler_arm64	319.864 MiB
✅	dogstatsd_deb_amd64	30.293 MiB
✅	dogstatsd_rpm_amd64	30.293 MiB
✅	dogstatsd_suse_amd64	30.293 MiB
✅	iot_agent_deb_armhf	44.652 MiB
✅	docker_cluster_agent_amd64	211.196 MiB
✅	docker_cluster_agent_arm64	224.224 MiB

[cit-pr-commenter-54b7da]

Regression Detector

Regression Detector Results

Metrics dashboard
Target profiles
Job ID: fd2afac7-f7ec-4c4b-a753-52cb135730be

Baseline: 1e25e03
Comparison: 41549ad
Diff

Optimization Goals: ✅ No significant changes detected

Fine details of change detection per experiment

perf	experiment	goal	Δ mean %	Δ mean % CI	trials	links
➖	dsd_uds_10mb_3k_timestamped_contexts_cpu	% cpu utilization	+0.76	[+0.49, +1.04]	1	Logs
➖	quality_gate_metrics_logs	memory utilization	+0.66	[+0.43, +0.89]	1	Logs bounds checks dashboard
➖	quality_gate_private_action_runner	memory utilization	+0.21	[+0.08, +0.33]	1	Logs bounds checks dashboard
➖	quality_gate_security_idle	memory utilization	+0.14	[+0.10, +0.18]	1	Logs bounds checks dashboard
➖	quality_gate_security_mean_fs_load	memory utilization	+0.04	[-0.00, +0.08]	1	Logs bounds checks dashboard
➖	quality_gate_security_no_fs_load	memory utilization	+0.03	[-0.05, +0.10]	1	Logs bounds checks dashboard
➖	quality_gate_idle_all_features	memory utilization	-0.01	[-0.09, +0.07]	1	Logs bounds checks dashboard
➖	quality_gate_idle	memory utilization	-0.06	[-0.10, -0.02]	1	Logs bounds checks dashboard
➖	dsd_uds_10mb_3k_timestamped_contexts_memory	memory utilization	-0.13	[-0.34, +0.09]	1	Logs
➖	python_openmetrics	% cpu utilization	-0.21	[-0.89, +0.47]	1	Logs bounds checks dashboard
➖	quality_gate_logs	% cpu utilization	-1.22	[-2.09, -0.35]	1	Logs bounds checks dashboard
➖	dsd_uds_client_drop_detector_cpu	% cpu utilization	-2.33	[-2.83, -1.83]	1	Logs

Bounds Checks: ✅ Passed

perf	experiment	bounds_check_name	replicates_passed	observed_value	links
✅	python_openmetrics	checks_execution_time	10/10	90.02 ≤ 100	bounds checks dashboard
✅	python_openmetrics	cpu_usage	10/10	1330.13 ≤ 1500	bounds checks dashboard
✅	python_openmetrics	memory_usage	10/10	4.33GiB ≤ 4.75GiB	bounds checks dashboard
✅	quality_gate_idle	intake_connections	10/10	4 ≤ 5	bounds checks dashboard
✅	quality_gate_idle	memory_usage	10/10	177.70MiB ≤ 181MiB	bounds checks dashboard
✅	quality_gate_idle	total_bytes_received	10/10	762.15KiB ≤ 819.20KiB	bounds checks dashboard
✅	quality_gate_idle_all_features	intake_connections	10/10	2 ≤ 5	bounds checks dashboard
✅	quality_gate_idle_all_features	memory_usage	10/10	475.04MiB ≤ 542MiB	bounds checks dashboard
✅	quality_gate_idle_all_features	total_bytes_received	10/10	1.14MiB ≤ 1.25MiB	bounds checks dashboard
✅	quality_gate_logs	intake_connections	10/10	17 ≤ 40	bounds checks dashboard
✅	quality_gate_logs	memory_usage	10/10	211.13MiB ≤ 228MiB	bounds checks dashboard
✅	quality_gate_logs	missed_bytes	10/10	0B = 0B	bounds checks dashboard
✅	quality_gate_logs	total_bytes_received	10/10	264.12MiB ≤ 292MiB	bounds checks dashboard
✅	quality_gate_metrics_logs	cpu_usage	10/10	422.77 ≤ 2000	bounds checks dashboard
✅	quality_gate_metrics_logs	intake_connections	10/10	19 ≤ 40	bounds checks dashboard
✅	quality_gate_metrics_logs	memory_usage	10/10	435.46MiB ≤ 455MiB	bounds checks dashboard
✅	quality_gate_metrics_logs	missed_bytes	10/10	0B = 0B	bounds checks dashboard
✅	quality_gate_metrics_logs	total_bytes_received	10/10	0.95GiB ≤ 1.04GiB	bounds checks dashboard
✅	quality_gate_private_action_runner	memory_usage	10/10	74.15MiB ≤ 77MiB	bounds checks dashboard
✅	quality_gate_security_idle	cpu_usage	10/10	35.25 ≤ 100	bounds checks dashboard
✅	quality_gate_security_idle	memory_usage	10/10	326.74MiB ≤ 357MiB	bounds checks dashboard
✅	quality_gate_security_mean_fs_load	cpu_usage	10/10	65.63 ≤ 200	bounds checks dashboard
✅	quality_gate_security_mean_fs_load	memory_usage	10/10	305.43MiB ≤ 337MiB	bounds checks dashboard
✅	quality_gate_security_no_fs_load	cpu_usage	10/10	29.56 ≤ 100	bounds checks dashboard
✅	quality_gate_security_no_fs_load	memory_usage	10/10	333.39MiB ≤ 348MiB	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_logs, bounds check missed_bytes: 10/10 replicas passed. Gate passed.
• quality_gate_logs, bounds check memory_usage: 10/10 replicas passed. Gate passed.
• quality_gate_logs, bounds check intake_connections: 10/10 replicas passed. Gate passed.
• quality_gate_logs, bounds check total_bytes_received: 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_metrics_logs, bounds check memory_usage: 10/10 replicas passed. Gate passed.
• quality_gate_metrics_logs, bounds check total_bytes_received: 10/10 replicas passed. Gate passed.
• quality_gate_idle_all_features, bounds check total_bytes_received: 10/10 replicas passed. Gate passed.
• quality_gate_idle_all_features, 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_security_mean_fs_load, bounds check memory_usage: 10/10 replicas passed. Gate passed.
• quality_gate_security_mean_fs_load, bounds check cpu_usage: 10/10 replicas passed. Gate passed.
• quality_gate_security_no_fs_load, bounds check memory_usage: 10/10 replicas passed. Gate passed.
• quality_gate_security_no_fs_load, bounds check cpu_usage: 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, bounds check total_bytes_received: 10/10 replicas passed. Gate passed.
• quality_gate_security_idle, bounds check memory_usage: 10/10 replicas passed. Gate passed.
• quality_gate_security_idle, bounds check cpu_usage: 10/10 replicas passed. Gate passed.
• quality_gate_private_action_runner, bounds check memory_usage: 10/10 replicas passed. Gate passed.

[atkinsam]

I think this would result in us removing the container for a healthy->unhealthy transition, even if the container is still running? Is that intentional?

[atkinsam]

Should we change this filter to use a single source? To avoid races between different sources emitting different events for the same container?