Google Cloud·24 Jan 2025 · 09:30 PSTResolved

Google AppSheet unavailable in two regions

A production database schema migration triggered cascading database failures and retry load, preventing many customers from loading AppSheet apps in two Google Cloud regions.

Reported duration1h 50m
Impact scopeRegional
Incident categoryDatabase / application
WHAT HAPPENED

Incident summary

A production database schema migration triggered cascading database failures and retry load, preventing many customers from loading AppSheet apps in two Google Cloud regions.

Approximately 60% of AppSheet requests in us-east4 and europe-west4 may have failed during the 1 hour and 50 minute incident. Affected users saw HTTP 500 errors, timeouts, and intermittent latency.

ROOT CAUSE

What the provider reported

Google Cloud reported that the migration caused failures and timeouts on the primary database. A surge of retries overloaded the secondary authentication database, and traffic migration then increased load on licensing services.

Editorial note: This is a concise summary of the provider’s public explanation, not an independent technical investigation.

AFFECTED SURFACE

Services and systems

AppSheetAppSheet editorApp load pageAuthentication

Reported scope: Northern Virginia and Netherlands

EVENT TIMELINE

Latest source updates

Showing up to the 3 latest updates available in the source record. UTC is fixed; local time is converted in this browser (your local timezone).

  1. 09:30 PST Local conversion unavailable: source time is not normalized

    Incident begins

    AppSheet starts returning errors and timeouts in us-east4 and europe-west4.

  2. 09:42 PST Local conversion unavailable: source time is not normalized

    Automated alert

    Google engineers are alerted and begin investigation.

  3. 11:20 PST Local conversion unavailable: source time is not normalized

    Incident ends

    Authentication is restored and traffic is gradually returned to the affected regions.

COMMUNITY SIGNAL

What did this incident feel like?

React anonymously. Counts are shared across visitors and are not a measure of the number of customers affected.

Community signal
Shared anonymous reactions