status page: curated incident record drives bars, uptime and Past Incidents; retire the dead gateway probe - #7
Merged
Conversation
…s; retire the dead gateway probe The page derived downtime from how long each incident ISSUE stayed open, which turned a 15-minute API blip whose issue lingered a day into ~1,440 minutes down (85.67% week) and kept a revoked-probe-key 401 on the books as a 30% 'gateway'. - assets/status-ui/incidents.json: 15 incidents reconstructed from the gateway ledger (per-10-minute completions / gateway-owned failures), production deploy runs and the day's incident notes, with the real UTC start/end and a short technical description each. 'down' counts against uptime; 'degraded' is shown yellow and not counted. Live checker issues after liveIssuesSince merge in; annulledIssues [4, 5] are the two monitoring artifacts. - uptime-bars.js renders the 90-day bars, legend uptime and the dated Past Incidents (component · state · HH:MM–HH:MM UTC · minutes + description) from that merged record. - .upptimerc.yml: Gateway (authenticated) disabled until a status-monitor key exists; the generator's issue-derived uptime figure and period selector are hidden so the page carries one uptime number. - README documents the record and how to add or annul an incident. Resulting 90-day uptime: API 99.83%, Web 99.78%, Docs 100%. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Why
The status page read "weird": a permanently red Gateway (authenticated) row at 30% uptime, an API incident that showed ~1,445 minutes for a blip, an 85.67% weekly figure, and bars that were only ever red or green.
Root causes: (1) Upptime derives downtime from how long each incident issue stayed open, not from the checks, so an issue that lingered a day counted as a day down; (2) the authenticated probe key was revoked on 2026-09-05 during a hand rotation that never updated the repo secret, so that row sent a dead key and reported a false
downfor two days; (3) nomaxResponseTimewas set, so the degraded (yellow) state was unreachable.Changes
assets/status-ui/incidents.json: 15 incidents reconstructed from the gateway ledger (completions and gateway-owned failures per 10-minute bucket), production deploy runs, and the day's incident notes, each with the real UTC start/end and a short technical description.down(not serving) counts against uptime;degraded(elevated errors/latency) renders yellow and does not. Live checker issues opened afterliveIssuesSincemerge in automatically;annulledIssues[4, 5]are the two monitoring artifacts.uptime-bars.jsnow renders the 90-day bars, the legend uptime, and the dated Past Incidents section (component · Outage/Degraded · HH:MM–HH:MM UTC · minutes, plus description) from that merged record..upptimerc.yml:maxResponseTime: 5000on every site (yellow is reachable); Gateway (authenticated) disabled until a key is minted under thestatus-monitororg (a key from any other org must never be used here); the generator's issue-derived uptime figure and its period selector are hidden so the page carries one uptime number.Resulting 90-day uptime: API 99.83%, Web 99.78%, Docs 100% (real downtime: API 33 min, Web 41 min).
Follow-up (after merge)
Remove the
statuslabel from issues #4 and #5 and close #4 so Upptime's own numbers stop counting them; dispatch Static Site CI so the new overlay deploys.🤖 Generated with Claude Code