rbxdata

All capture systems operational

124 ok · 0 pending · fetched just now

alarm history · last 14 days · 3 of 84 intervals saw an alarm
Oct 2Oct 4Oct 6Oct 8Oct 10Oct 12Oct 14now

Alarms ok

124

Pending

0

Firing

0

Live events

2,375

worklist atoms · boundary 6,000

Sweeper peak

251 MB

of 384 MB · alarm at 345

Worklist pressure

80.2%

worst: recommendations-daily · alarm at 85%

Snapshot

15 h 43 min ago

data day 2026-10-15

Capture throughput

raw objects written per source · last 90d

games

87,376

virtual-events

140,026

age-rating

69,419

badges

222,447

Live-events worklist

atoms per freshness pass · boundary 6,000 (S3.10) · pages via the fleet max-ratio alarm at 85%

Shared sweeper memory

peak MB per invocation · 384 MB function (S3.10) · alarm at 345

Capture freshness

one alarm per capture job (coverage-record attribution, M-4/C-15)

Worklists

compile staleness + lane-ceiling approach

Memory headroom

90% of each function's memory (the F-15 growth-exhaustion class)

Canaries, pipeline & cost

everything else carrying the rbxdata- prefix

Ops surface: reads CloudWatch on every visit; noindexed. Rows open each alarm's history; the capture-side twin is /coverage.