[AGNTLOG-604] Protect important logs from adaptive sampling

[gh123man]

Summary

• Extends the tokenizer with 13 critical severity token types (FATAL, ERROR, PANIC, WARN, CRITICAL, ALERT, SEVERE, EMERGENCY, EXCEPTION, CRASH, FAILURE, DEADLOCK, TIMEOUT)
• Adds a bypass in AdaptiveSampler.Process() that protects logs containing these tokens from being dropped — critical keywords are promoted during the existing tokenization pass (single-pass, no extra byte scan or regex)
• New config: logs_config.experimental_adaptive_sampling.protect_important_logs (default: true)

Motivation

The adaptive sampler rate-limits repetitive log patterns, but has no concept of importance — a FATAL: disk full gets the same treatment as [DEBUG] cache hit. This change ensures critical logs are never dropped regardless of pattern frequency.

Test plan

• 156 unit tests passing (tokenizer + sampler)
• 24 new tokenizer test cases: keyword promotion, case insensitivity, false-positive safety
• 5 new sampler tests: bypass works, no entry creation, non-important still dropped, protect disabled, isImportant unit
• New benchmark: BenchmarkSampler_Adaptive_ImportantBypass
• CI pipeline validates build + lint

🤖 Generated with Claude Code

[dd-octo-sts]

Files inventory check summary

File checks results against ancestor a4cc4921:

Results for datadog-agent_7.79.0~devel.git.689.45e1def.pipeline.107480490-1_amd64.deb:

No change detected

[cit-pr-commenter-54b7da]

Regression Detector

Regression Detector Results

Metrics dashboard
Target profiles
Run ID: 140a8ab5-6700-440b-a964-52036504ae6e

Baseline: 0a87ca0
Comparison: 817dfda
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	+3.36	[+0.28, +6.44]	1	Logs

Fine details of change detection per experiment

perf	experiment	goal	Δ mean %	Δ mean % CI	trials	links
➖	docker_containers_cpu	% cpu utilization	+3.36	[+0.28, +6.44]	1	Logs
➖	ddot_logs	memory utilization	+1.24	[+1.17, +1.32]	1	Logs
➖	ddot_metrics	memory utilization	+0.72	[+0.53, +0.91]	1	Logs
➖	ddot_metrics_sum_delta	memory utilization	+0.47	[+0.30, +0.64]	1	Logs
➖	ddot_metrics_sum_cumulativetodelta_exporter	memory utilization	+0.11	[-0.12, +0.34]	1	Logs
➖	file_to_blackhole_1000ms_latency	egress throughput	+0.07	[-0.36, +0.50]	1	Logs
➖	file_to_blackhole_500ms_latency	egress throughput	+0.05	[-0.36, +0.45]	1	Logs
➖	uds_dogstatsd_to_api_v3	ingress throughput	+0.02	[-0.20, +0.24]	1	Logs
➖	tcp_dd_logs_filter_exclude	ingress throughput	+0.01	[-0.10, +0.11]	1	Logs
➖	uds_dogstatsd_to_api	ingress throughput	+0.00	[-0.20, +0.21]	1	Logs
➖	otlp_ingest_metrics	memory utilization	-0.01	[-0.17, +0.15]	1	Logs
➖	file_to_blackhole_100ms_latency	egress throughput	-0.04	[-0.15, +0.08]	1	Logs
➖	ddot_metrics_sum_cumulative	memory utilization	-0.09	[-0.23, +0.05]	1	Logs
➖	file_to_blackhole_0ms_latency	egress throughput	-0.15	[-0.66, +0.37]	1	Logs
➖	quality_gate_idle	memory utilization	-0.26	[-0.31, -0.21]	1	Logs bounds checks dashboard
➖	uds_dogstatsd_20mb_12k_contexts_20_senders	memory utilization	-0.29	[-0.36, -0.22]	1	Logs
➖	docker_containers_memory	memory utilization	-0.32	[-0.40, -0.24]	1	Logs
➖	quality_gate_idle_all_features	memory utilization	-0.36	[-0.40, -0.32]	1	Logs bounds checks dashboard
➖	file_tree	memory utilization	-0.42	[-0.47, -0.36]	1	Logs
➖	tcp_syslog_to_blackhole	ingress throughput	-0.70	[-0.86, -0.54]	1	Logs
➖	quality_gate_logs	% cpu utilization	-0.88	[-2.51, +0.75]	1	Logs bounds checks dashboard
➖	quality_gate_metrics_logs	memory utilization	-1.01	[-1.25, -0.77]	1	Logs bounds checks dashboard
➖	otlp_ingest_logs	memory utilization	-1.04	[-1.14, -0.93]	1	Logs

Bounds Checks: ✅ Passed

perf	experiment	bounds_check_name	replicates_passed	observed_value	links
✅	docker_containers_cpu	simple_check_run	10/10	683 ≥ 26
✅	docker_containers_memory	memory_usage	10/10	275.78MiB ≤ 370MiB
✅	docker_containers_memory	simple_check_run	10/10	695 ≥ 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	0.20GiB ≤ 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	175.53MiB ≤ 181MiB	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	497.58MiB ≤ 550MiB	bounds checks dashboard
✅	quality_gate_logs	intake_connections	10/10	4 ≤ 6	bounds checks dashboard
✅	quality_gate_logs	memory_usage	10/10	203.61MiB ≤ 220MiB	bounds checks dashboard
✅	quality_gate_logs	missed_bytes	10/10	0B = 0B	bounds checks dashboard
✅	quality_gate_metrics_logs	cpu_usage	10/10	342.60 ≤ 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	417.17MiB ≤ 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_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_metrics_logs, bounds check cpu_usage: 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 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_logs, 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.

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 0f2d9cc67c

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".

[gh123man]

/merge

[gh-worker-devflow-routing-ef8351]

View all feedbacks in Devflow UI.

2026-04-14 17:28:44 UTC ℹ️ Start processing command /merge

---

2026-04-14 17:28:49 UTC ℹ️ MergeQueue: pull request added to the queue

The expected merge time in main is approximately 5h (p90).

---

2026-04-14 18:35:56 UTC ℹ️ MergeQueue: Readding this merge request to the queue because another merge request processed with yours failed. No action is needed from your side.

---

2026-04-14 20:17:44 UTC ℹ️ MergeQueue: Readding this merge request to the queue because another merge request processed with yours failed. No action is needed from your side.

---

2026-04-14 21:09:34 UTC ℹ️ MergeQueue: This merge request was merged

[soberpeach]

Approving with a question for now.

Do you think it would be worth it to check for Protect logs before the sampler? Perhaps in the combiner? This way we could bypass the traversing the tokens again in the sampler. We could also just add a field onto the log while it's being tokenized and then use that information to rate limit it or not.