fix: correct competitive bench line counting for OTLP and NDJSON

[strawgate]

Summary

• fix blackhole line counting used by competitive benchmarks
• count OTLP /v1/logs records by parsing resourceLogs[].scopeLogs[].logRecords[]
• keep fallback to "body": scan when OTLP JSON parsing is unavailable
• fix NDJSON counting for payloads without trailing newline
• add regression tests for NDJSON, ES bulk, and OTLP log record counting

Why

Benchmark throughput can be under-reported when payloads do not end with \n (notably affecting jsonline senders like vlagent), and OTLP payload counting should use log-record semantics rather than newline heuristics.

Validation

• cargo test -p logfwd-competitive-bench
• cargo check -p logfwd-competitive-bench
• just ci in this environment is blocked by pre-existing cargo deny duplicate warnings unrelated to this patch

Notes

I attempted focused docker benchmark runs for otelcol and vlagent, but benchmark execution is blocked/hangs in this environment. The code path and regression tests are in place for deterministic counting behavior.

[coderabbitai]

Caution

Review failed

Pull request was closed or merged during review

Walkthrough

The blackhole benchmark's request counting for /v1/logs now tries OTLP JSON structural counting by parsing the body and summing resourceLogs[].scopeLogs[].logRecords[]. If parsing yields no count it falls back to the prior substring "body:" occurrence scan, and if that still yields zero it uses an NDJSON-style estimator (ndjson_line_count) which returns 0 for empty bodies and adds 1 when the body lacks a trailing newline. NDJSON/newline handling and OTLP counting unit tests were added.

Possibly related PRs

• feat: Benchmark reorg — fix failures, dual mode, github-action-benchmark #66: Modifies the same blackhole counting logic to add OTLP JSON counting and NDJSON/newline fallback behavior.
• feat: Benchmark logfwd vs vector, fluent-bit, filebeat, otelcol, and vlagent #33: Adjusts bench/run behavior to rely on parsed lines from the blackhole, closely coupled to the updated counting output.

---

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

[coderabbitai]

Actionable comments posted: 1

🤖 Prompt for all review comments with AI agents

Verify each finding against the current code and only fix it if needed.

Inline comments:
In `@crates/logfwd-competitive-bench/src/blackhole.rs`:
- Around line 109-116: The OTLP counting currently calls count_otlp_log_records
and count_otlp_body_keys unconditionally; change to lazy-evaluate the fallback
so you only run the heavier substring/scan when the JSON parse returned no
records: first call count_otlp_log_records(body) and if it returns a positive
count return it immediately; only if it returns 0 then call
count_otlp_body_keys(body) and return that if >0; otherwise fall back to
ndjson_line_count(body, newlines). Update the logic around the otlp_count/return
path (the block guarded by url.contains("/v1/logs")) to reflect this
short-circuit behavior.

🪄 Autofix (Beta)

Fix all unresolved CodeRabbit comments on this PR:

• Push a commit to this branch (recommended)
• Create a new PR with the fixes

---

ℹ️ Review info

⚙️ Run configuration

Configuration used: Organization UI

Review profile: ASSERTIVE

Plan: Pro

Run ID: f33bedf7-c108-4817-84e2-5c993dbb012a

📥 Commits

Reviewing files that changed from the base of the PR and between 6380b91 and 3c17fe5.

📒 Files selected for processing (1)

• crates/logfwd-competitive-bench/src/blackhole.rs