Critical

Coframe scripts stopped firing after theme publish

updated 0s ago
State
Firing
unassigned · never acknowledged
Open for
2h 14m
first seen 09:42 UTC · Jul 23
Blast radius
12stores
4.1% of fleet sessions · 1.72M/day
Collapsed from
38alerts
rule: Script fire-rate floor

Script fire rate · affected stores

threshold 95% for 10m
00:0006:00 09:42 breach now 11:56
Fell from 99.1% to 78.4% in 4 minutes — a step change, not a drift. Consistent with a deploy, not with traffic.

Affected stores

12 of 1,284 · sorted by sessions lost
StoreSDKTheme publishedFire rate Sessions lost / day
northwind-supply 4.2.1 09:38 · Dawn 15.2.0 0% 412,900
atlas-outfitters 4.2.1 09:41 · Dawn 15.2.0 0% 338,140
harbor-lane 4.2.1 09:44 · Craft 6.1.1 31% 201,470
foundry-mercantile 4.2.1 09:47 · Dawn 15.2.0 0% 186,020
pilot-goods 4.2.1 10:02 · Dawn 15.2.0 44% 154,880
kettle-and-co 4.2.1 10:16 · Refresh 9.0.4 0% 129,310
6 more stores below 100k sessions/day

Timeline

all times UTC
Jul 22 · 14:00
SDK 4.2.1 rollout reached 62% of fleet
Automatic staged rollout. No alert at the time — coverage looked normal for 19 hours.
Jul 23 · 09:38
northwind-supply published theme Dawn 15.2.0
First of 12 stores. Publish events observed on 11 more within 38 minutes.
Jul 23 · 09:42
Rule Script fire-rate floor breached
Fire rate 78.4% against a 95% floor, sustained 10 minutes. INC-2291 opened.
Jul 23 · 09:44
37 further alerts collapsed into this incident
Same rule, same SDK version, same 38-minute window — grouped instead of paged separately.
Jul 23 · 10:07
Paged on-call · no acknowledgement
Escalation policy scripts-critical · re-page due at 12:07.
Jul 23 · 11:56
Still firing · 12 stores
No change in blast radius for 74 minutes.

Correlation

All 12 stores run SDK 4.2.1 and published a theme after 09:30. No store on 4.2.0 is affected, and no 4.2.1 store that skipped a publish is affected.

Shared SDK
4.2.1 (12 / 12)
Shared action
Theme publish (12 / 12)
Shared theme
Dawn 15.2.0 (9 / 12)
Shared region
none — 4 regions
Shared plan
none — 3 plans
Confidence: two signals at 12/12

Runbook

scripts-silent
  1. Confirm the script tag survived the publish on one affected store.
  2. If absent, re-run Install Coframe Scripts for the store from the admin API.
  3. If present but inert, roll the store back to SDK 4.2.0 and re-check fire rate.
  4. Hold the 4.2.1 rollout until fire rate recovers above 95% fleet-wide.
coframe scripts:verify \ --store northwind-supply \ --expect-version 4.2.1

Seen before

Both fixed by re-running script install.

Prototype · figures are sample data for layout review, not production telemetry.