posthog/diagnosing-ci-and-merge-bottlenecks

Diagnosing CI and merge bottlenecks

Engineering analytics treats a pull request like product analytics treats a user: a PR moves through a pipeline (opened → CI → review → merged → deployed) and the job is to find where it slows down. The surface is three named MCP tools — you call them, you don't write SQL. Dogfooded on PostHog/posthog; the same tools serve autonomous agents (e.g. PostHog Code) reasoning about their own PRs.

The tools

• pull-requests — the PR workhorse. Open PRs plus anything merged or closed since date_from (default -30d), newest first. Each row carries author (nested object: handle, display_name, is_bot), repo (nested: owner, name), state, is_draft, labels, open_to_merge_seconds, and a ci rollup (runs / passing / failing / pending) from the head-SHA join. Answers most PR-level questions: which PRs have failing or pending CI, which are stuck open longest, per-author or per-repo triage, and time-to-merge stats (aggregate open_to_merge_seconds over the returned merged rows yourself — median and p95, never a mean).
• workflow-health — per-workflow CI health over a window (date_from / date_to, default last 30 days): run_count, success_rate, p50_seconds, p95_seconds, last_failure_at. Answers "is CI getting faster or slower" and "which workflow is the slow or flaky long pole". There is no built-in trend — call it over two adjacent windows and compare. success_rate / p50_seconds / p95_seconds cover completed runs only and are null when a window has no completed runs — guard for null before comparing two windows (a workflow can have runs in one and none in the other).
• pr-lifecycle — a single PR's timeline: a header plus ordered events — opened, then a CI started/finished pair per workflow run (many on a multi-workflow repo, interleaved by time), then merged/closed. Answers "where is PR N stuck". metric_quality is partial.

There is no aggregate time-to-merge tool and no "counts" tool — derive those from pull-requests (the stuck/failing counts, the merge-time percentiles).

Caveats you must carry into every answer

These are structural limits of today's snapshot data — state them, don't paper over them.

• open_to_merge_seconds is coarse. It fuses draft time and ready-for-review time into one figure. Report it as "open to merge", never "cycle time" or "review time". Flag it when long-lived drafts inflate a number.
• CI status can be stale. The CI source syncs on a watermark and does not refresh a run that completes after newer runs land (until the workflow_run webhook ships). Treat a pending count as unsettled, not as a settled failure; lead with status, not a verdict.
• CI for a PR is the head-SHA join, nothing else. The ci rollup reflects only the latest commit's runs. There is no other link between a PR and its checks.
• No reviews, approvals, per-check/job, or deploys yet. Don't infer review behaviour or DORA metrics from their absence; that data hasn't landed. pr-lifecycle is partial for the same reason.
• Bots and drafts are present in pull-requests output, excluded by convention. Filter out author.is_bot (nested under author, not a row-level field) and is_draft for throughput / merge-time questions; keep them in for bot-impact questions.
• pull-requests returns a capped page. At most limit rows (newest first); truncated is true when more match, and there is no repo or limit filter to narrow the call. When truncated is true, any percentile or count you derive covers only the newest page — not the whole window — so say so and shrink date_from until the real set fits under the cap.

Choosing a tool

| The question | Tool | How | | ------------------------------------------------------ | ----------------- | ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | Is CI getting slower? Which workflow is the long pole? | workflow-health | Call over two adjacent windows (e.g. date_from=-14d, then date_from=-28d date_to=-14d); compare p50_seconds and p95_seconds per workflow. Lead with the median but always check p95 separately — they move independently. | | Which open PRs have failing or pending CI? | pull-requests | Keep rows where ci.failing > 0 or ci.pending > 0. pending means unsettled (or stale) — not a settled failure. | | Which PRs are stuck open longest? | pull-requests | Keep state = open, not is_draft, not author.is_bot; sort by created_at ascending (oldest first). | | How long are PRs taking to merge? Per author? | pull-requests | Over merged rows (merged_at set, not bot, not draft), aggregate open_to_merge_seconds — median and p95. Group by author.handle for cohort context, not a ranking (per-developer surveillance is an explicit non-goal). Trend it by calling with two date_from windows. | | Where is PR N stuck? | pr-lifecycle | Walk the sorted events: opened → first CI started, the CI span (first start → last finish; one pair per workflow), last CI finished → merged. The largest gap is the bottleneck. A long open→merge with quick CI points at review/idle time the partial data can't itemize yet — say so. |

The high-value chain

Mirror how a human investigates: aggregate signal → confirm → concrete PR.

workflow-health  (find the slow/flaky long-pole workflow)
   → pull-requests  (confirm it's dragging merge time; list the affected PRs)
      → pr-lifecycle  (open a representative stuck PR and show the gap)

"CI median rose because e2e-playwright p95 doubled; that workflow is the long pole on PR #1234, which sat 47m in CI before merging."

Output expectations

• Lead with the verdict in one line, then the supporting numbers.
• Carry the coarse / partial / staleness caveat whenever the distinction matters.
• For multi-window or multi-workflow comparisons, a short table beats prose. Report median and p95 side by side — never collapse them into one "average".

What NOT to do

• Don't call open_to_merge_seconds cycle time or review time — it's coarse open-to-merge.
• Don't report a CI count as a settled failure when pending > 0 — it may be unsettled or stale.
• Don't infer reviews, approvals, per-check counts, or deploys — that data isn't ingested yet.
• Don't turn per-author buckets into a leaderboard — they're for finding stuck work, not ranking people.
• Don't reach for these tools to fetch raw PR contents or diffs — they surface pipeline signal, not the PR thread.