Elevated API Errors on App Platform upto 7 %

Incident Report for Zeotap

Postmortem

Incident window (IST)

15 May 2025 — 11:59 to 12:26 UTC (≈ 27 minutes)

Service impacted

app.zeotap.com dashboards and APIs (partial outage, global)

What happened

A newly-deployed Cloudflare rule—meant to route a specific marketing-widget path to a new backend—was too broadly scoped. Legitimate UI traffic was unintentionally rewritten and forwarded to an inactive origin, resulting in 4xx/5xx errors.

Customer impact

  • UI pages failed to load certain resources, leading to blank or error states.
  • Approx. 2 k requests returned errors during the 27-minute window

Root cause

A logical grouping mistake in the rule’s filter (AND / OR precedence) caused it to match all requests that contained particular referrer domains, regardless of host or path.

Immediate remediation

  • 12:26 — Rule disabled and cache purged.
  • Error rates returned to normal within seconds.

Preventive actions

  1. Safer change process – Cloudflare rules will move to Infrastructure-as-Code with mandatory peer review.
  2. Staged roll-outs – Future edge-rules will deploy first to a staging zone and a limited canary cohort.
  3. Faster detection – Synthetic checks for critical UI paths will alert after three consecutive failures (≈ 3 min).

We apologise for the disruption and appreciate your patience while we improve our safeguards.

Posted May 16, 2025 - 07:32 UTC

Resolved

Between 12:03 UTC and 12:28 UTC we saw an elevated API error rate (peaking at ≈ 7 %). The underlying issue has been fixed and the platform is operating normally. We are monitoring closely.
Posted May 15, 2025 - 12:30 UTC