[ AGNTLOG-703 ] Fix syslog TCP frame misclassification and recognize appliance timestamp layouts

[ddrthall]

What does this PR do?

Fixes two classes of defect in syslog collection over sockets, and adds a regression harness for the second.

The TCP framer no longer commits to RFC 6587 octet counting on the strength of a leading digit. It requires the whole MSG-LEN SP PRI signature — with a complete PRI, meaning <, a one-to-three digit PRIVAL, then > — before any digits are read as a length, and falls back to newline recovery otherwise. Because a TCP read can split anywhere, the test is three-way: an incomplete signature waits for more bytes rather than guessing. Frame detection and malformed-run resynchronization now share that one test, so the two cannot disagree about where a frame starts.

The parser recognizes the header layouts appliances actually emit, and stops putting values in the wrong fields. Newly handled: non-padded single-digit days, Cisco NX-OS year-first timestamps, a colon terminating the PRI as in Cisco's EMBLEM dialect, ISO 8601 in place of the BSD timestamp, and the Cisco IOS service timestamps options for fractional seconds, timezone names, and the unsynchronized-clock marker. Newly corrected: a TAG sitting in the HOSTNAME position, a relay's second timestamp mined for an APP-NAME, padded RFC 5424 field separators, an unterminated TAG swallowing the body, and the separator Cisco leaves behind when the Device-ID is disabled.

Alongside those, testdata/corpus.json pins the parser's complete output for 75 messages across 22 integrations, spanning all seven framing classes and both the wire form and the PRI-less rendering a tailed file carries.

Motivation

The framing bug lost data silently and without bound, but only behind a payload that was already malformed. Framing method is chosen at a frame boundary, and conformant traffic opens every frame with a PRI, so reaching the digit-leading path at all requires the payload sitting at that boundary to be non-conformant. From there MSG-LEN is the authoritative frame boundary and the body it declares is never re-scanned for frame starts, so a digit run misread as a length swallowed everything behind it into one truncated log. The messages it swallowed could themselves be perfectly well-formed, which is what turned a single malformed payload into unbounded loss.

The parser defects degraded attribution rather than content. An unrecognized layout returns every header field as the nilvalue, which alone makes source remapping on syslog.appname or syslog.hostname impossible. A misattributed field is worse, because those same attributes drive source and service routing, so the log is misrouted with nothing to signal that anything went wrong — a timezone name read as an APP-NAME would route every device in a zone under a service named after it. The original line is preserved throughout; only extracted attributes change.

The corpus exists because targeted tests cannot catch that second class. One sample per parsing decision is the right shape for asserting a behaviour, but nobody writes an assertion that HOSTNAME is not network:, so nothing fails when it is. Comparing every field of every message turns a quiet misattribution into a diff.

Every accepted layout is corroborated by vendor documentation or an independent parser implementation rather than by a single fixture; one tolerance that could not be reproduced beyond its originating fixture was dropped rather than shipped. Cisco's documented collector expectations drove the most recent round: the guides define the IOS header as seq no:timestamp: %facility-severity-MNEMONIC:description and enumerate the service timestamps options that vary everything ahead of the %.

Describe how you validated your changes

dda inv test --targets=./pkg/logs/internal/framer,./pkg/logs/internal/parsers/syslog, also run under -race.

The golden corpus is the primary check on the parser work: reverting any one of the fixes makes it fail and name the integration, the fixture line, and the field that moved. Each framer regression test was run against the unfixed framer to confirm it reproduces the corruption rather than passing vacuously. Both parser entry points were fuzzed with no crashes, and benchstat reports the combined parser changes as throughput-neutral with allocations per operation unchanged.

New coverage sits in syslog_octet_prefix_test.go, which asserts that a genuine octet-counted frame is still framed by length, that digit-leading non-frames fall through to newline recovery instead of swallowing their successors, and that a length prefix straddling a read boundary is not misclassified; and in timestamp_variants_test.go, which covers each accepted layout and cites the fixture or vendor guide it came from, noting whether the sample is a wire capture or a file rendering since only the former carries a PRI.

Additional Notes

Two layouts Cisco documents are deliberately left unhandled. Uptime timestamps (00:00:46:, 1w2d:) are time since reboot rather than wall clock, so there is nothing to extract and passing the line through as content is already correct. The IOS sequence-number prefix is not yet recognized and remains a follow-up.

Repeated BSD headers are also out of scope here. Unlike the other layouts, a doubled BSD header is not something any integration documents its devices as emitting — it comes from a relay that fails to recognize an already-formatted line — so it is a separate problem with a separate scope. The ISO form of that check is untouched.

Every corpus message names its origin, and each origin records the public repository and pinned commit, the vendor document, or that the sample was synthesized, so any entry can be traced without guesswork. Regenerate after an intentional parser change with:

go test ./pkg/logs/internal/parsers/syslog -run TestCorpusGolden -update-golden

[datadog-prod-us1-4]

🎯 Code Coverage (details)
• Patch Coverage: 93.45%
• Overall Coverage: 55.83%

_{This comment will be updated automatically if new data arrives.
🔗 Commit SHA: 518c44c | Docs | Datadog PR Page | Give us feedback!}

[dd-octo-sts]

Files inventory check summary

File checks results against ancestor 36808442:

Results for datadog-agent_7.84.0~devel.git.24.385dbe3.pipeline.129724895-1_amd64.deb:

No change detected

Results for datadog-iot-agent_7.84.0~devel.git.24.385dbe3.pipeline.129724895-1_amd64.deb:

No change detected

[dd-octo-sts]

Static quality checks

✅ Please find below the results from static quality gates
Comparison made with ancestor f57e7a3
📊 Static Quality Gates Dashboard
🔗 SQG Job
SOME SIZE DELTAS ARE N/A (ANCESTOR METRICS NOT YET AVAILABLE). RETRY JOB

Successful checks

Info

	Quality gate	Change	Size (prev → curr → max)
✅	agent_deb_amd64	N/A	N/A → 759.550 → 764.430
✅	agent_deb_amd64_fips	N/A	N/A → 712.892 → 715.680
✅	agent_heroku_amd64	N/A	N/A → 312.277 → 318.990
✅	agent_msi	N/A	N/A → 643.255 → 660.900
✅	agent_rpm_amd64	N/A	N/A → 759.534 → 764.400
✅	agent_rpm_amd64_fips	N/A	N/A → 712.876 → 715.680
✅	agent_rpm_arm64	N/A	N/A → 735.503 → 736.490
✅	agent_rpm_arm64_fips	N/A	N/A → 692.108 → 693.640
✅	agent_suse_amd64	N/A	N/A → 759.534 → 764.400
✅	agent_suse_amd64_fips	N/A	N/A → 712.876 → 715.680
✅	agent_suse_arm64	N/A	N/A → 735.503 → 736.490
✅	agent_suse_arm64_fips	N/A	N/A → 692.108 → 693.640
✅	docker_agent_amd64	N/A	N/A → 818.162 → 819.970
✅	docker_agent_arm64	N/A	N/A → 819.174 → 820.690
✅	docker_agent_jmx_amd64	N/A	N/A → 1009.059 → 1010.730
✅	docker_agent_jmx_arm64	N/A	N/A → 998.724 → 1000.370
✅	docker_cluster_agent_amd64	N/A	N/A → 210.228 → 211.300
✅	docker_cluster_agent_arm64	N/A	N/A → 223.267 → 223.810
✅	docker_cws_instrumentation_amd64	N/A	N/A → 7.439 → 7.480
✅	docker_cws_instrumentation_arm64	N/A	N/A → 6.877 → 7.110
✅	docker_dogstatsd_amd64	N/A	N/A → 39.479 → 39.960
✅	docker_dogstatsd_arm64	N/A	N/A → 37.560 → 38.270
✅	docker_host_profiler_amd64	N/A	N/A → 305.801 → 317.690
✅	docker_host_profiler_arm64	N/A	N/A → 317.116 → 328.970
✅	dogstatsd_deb_amd64	N/A	N/A → 30.224 → 31.200
✅	dogstatsd_deb_arm64	N/A	N/A → 28.247 → 29.580
✅	dogstatsd_rpm_amd64	N/A	N/A → 30.224 → 31.200
✅	dogstatsd_suse_amd64	N/A	N/A → 30.224 → 31.200
✅	iot_agent_deb_amd64	N/A	N/A → 46.424 → 46.440
✅	iot_agent_deb_arm64	N/A	N/A → 43.099 → 43.770
✅	iot_agent_deb_armhf	N/A	N/A → 43.879 → 44.020
✅	iot_agent_rpm_amd64	N/A	N/A → 46.424 → 46.440
✅	iot_agent_suse_amd64	N/A	N/A → 46.423 → 46.440

[cit-pr-commenter-54b7da]

Regression Detector

Regression Detector Results

Metrics dashboard
Target profiles
Run ID: 65072894-045c-43ae-86c9-280f53c6a91b

Baseline: b39d95d
Comparison: 518c44c
Diff

Optimization Goals: ✅ No significant changes detected

Fine details of change detection per experiment

perf	experiment	goal	Δ mean %	Δ mean % CI	trials	links
➖	quality_gate_private_action_runner	memory utilization	+0.31	[+0.19, +0.43]	1	Logs bounds checks dashboard
➖	quality_gate_security_mean_fs_load	memory utilization	+0.13	[+0.10, +0.17]	1	Logs bounds checks dashboard
➖	quality_gate_idle	memory utilization	+0.01	[-0.05, +0.06]	1	Logs bounds checks dashboard
➖	quality_gate_security_idle	memory utilization	+0.00	[-0.05, +0.06]	1	Logs bounds checks dashboard
➖	quality_gate_idle_all_features	memory utilization	-0.03	[-0.06, +0.01]	1	Logs bounds checks dashboard
➖	quality_gate_security_no_fs_load	memory utilization	-0.13	[-0.22, -0.03]	1	Logs bounds checks dashboard
➖	quality_gate_logs	% cpu utilization	-0.42	[-1.27, +0.43]	1	Logs bounds checks dashboard
➖	quality_gate_metrics_logs	memory utilization	-1.02	[-1.26, -0.77]	1	Logs bounds checks dashboard

Bounds Checks: ✅ Passed

perf	experiment	bounds_check_name	replicates_passed	observed_value	links
✅	quality_gate_idle	intake_connections	10/10	3 ≤ 4	bounds checks dashboard
✅	quality_gate_idle	memory_usage	10/10	149.90MiB ≤ 154MiB	bounds checks dashboard
✅	quality_gate_idle	total_bytes_received	10/10	742.92KiB ≤ 819.20KiB	bounds checks dashboard
✅	quality_gate_idle_all_features	intake_connections	10/10	3 ≤ 4	bounds checks dashboard
✅	quality_gate_idle_all_features	memory_usage	10/10	493.80MiB ≤ 512MiB	bounds checks dashboard
✅	quality_gate_idle_all_features	total_bytes_received	10/10	1.12MiB ≤ 1.25MiB	bounds checks dashboard
✅	quality_gate_logs	intake_connections	10/10	17 ≤ 40	bounds checks dashboard
✅	quality_gate_logs	memory_usage	10/10	184.60MiB ≤ 195MiB	bounds checks dashboard
✅	quality_gate_logs	missed_bytes	10/10	0B = 0B	bounds checks dashboard
✅	quality_gate_logs	total_bytes_received	10/10	263.82MiB ≤ 292MiB	bounds checks dashboard
✅	quality_gate_metrics_logs	cpu_usage	10/10	361.48 ≤ 2000	bounds checks dashboard
✅	quality_gate_metrics_logs	intake_connections	10/10	18 ≤ 40	bounds checks dashboard
✅	quality_gate_metrics_logs	memory_usage	10/10	374.15MiB ≤ 430MiB	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	0.94GiB ≤ 1.04GiB	bounds checks dashboard
✅	quality_gate_private_action_runner	memory_usage	10/10	72.07MiB ≤ 75MiB	bounds checks dashboard
✅	quality_gate_security_idle	cpu_usage	10/10	27.91 ≤ 100	bounds checks dashboard
✅	quality_gate_security_idle	memory_usage	10/10	300.61MiB ≤ 330MiB	bounds checks dashboard
✅	quality_gate_security_mean_fs_load	cpu_usage	10/10	59.10 ≤ 200	bounds checks dashboard
✅	quality_gate_security_mean_fs_load	memory_usage	10/10	278.86MiB ≤ 310MiB	bounds checks dashboard
✅	quality_gate_security_no_fs_load	cpu_usage	10/10	21.42 ≤ 100	bounds checks dashboard
✅	quality_gate_security_no_fs_load	memory_usage	10/10	284.84MiB ≤ 320MiB	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".

Replicate Execution Details

We run multiple replicates for each experiment/variant. However, we allow replicates to be automatically retried if there are any failures, up to 8 times, at which point the replicate is marked dead and we are unable to run analysis for the entire experiment. We call each of these attempts at running replicates a replicate execution. This section lists all replicate executions that failed due to the target crashing or being oom killed.

Note: In the below tables we bucket failures by experiment, variant, and failure type. For each of these buckets we list out the replicate indexes that failed with an annotation signifying how many times said replicate failed with the given failure mode. In the below example the baseline variant of the experiment named experiment_with_failures had two replicates that failed by oom kills. Replicate 0, which failed 8 executions, and replicate 1 which failed 6 executions, all with the same failure mode.

Experiment	Variant	Replicates	Failure	Logs	Debug Dashboard
experiment_with_failures	baseline	0 (x8) 1 (x6)	Oom killed		Debug Dashboard

The debug dashboard links will take you to a debugging dashboard specifically designed to investigate replicate execution failures.

❌ Retried Profiling Replicate Execution Failures (ddprof)

Note: Profiling replicas may still be executing. See the debug dashboard for up to date status.

Experiment	Variant	Replicates	Failure	Debug Dashboard
quality_gate_idle	baseline	10	Oom killed	Debug Dashboard
quality_gate_idle_all_features	baseline	10	Oom killed	Debug Dashboard
quality_gate_idle_all_features	comparison	10	Oom killed	Debug Dashboard
quality_gate_logs	baseline	10	Oom killed	Debug Dashboard
quality_gate_logs	comparison	10	Oom killed	Debug Dashboard
quality_gate_metrics_logs	baseline	10	Oom killed	Debug Dashboard
quality_gate_metrics_logs	comparison	10	Oom killed	Debug Dashboard
quality_gate_security_idle	baseline	10	Oom killed	Debug Dashboard
quality_gate_security_no_fs_load	baseline	10	Crashed (exit code: 134)	Debug Dashboard
quality_gate_security_no_fs_load	comparison	10	Oom killed	Debug Dashboard

CI Pass/Fail Decision

✅ Passed. All Quality Gates 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_logs, bounds check memory_usage: 10/10 replicas passed. Gate passed.
• quality_gate_logs, bounds check total_bytes_received: 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.
• quality_gate_private_action_runner, 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 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 total_bytes_received: 10/10 replicas passed. Gate passed.
• quality_gate_idle, bounds check total_bytes_received: 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 intake_connections: 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 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.

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 54e51196d4

ℹ️ About Codex in GitHub

Codex has been enabled to automatically 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 👍.

When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".

[chatgpt-codex-connector]

Document the golden update with dda inv

This new regeneration hint tells contributors to run raw go test, but this repository’s Go packages rely on custom build tags and the repo-wide workflow requires dda inv test instead. In the documented update path for the golden corpus, following this command can produce misleading local results or bypass the intended test wrapper; please replace it with the dda inv test --targets=... equivalent.

AGENTS.md reference: AGENTS.md:L35-L46

Useful? React with 👍 / 👎.