Search outage with partial website degradation due to distributed scraping incident
Resolved
Oct 29, 2026 at 11:57pm UTC
On 30 October 2026, around 00:53–00:57 CET (29 October, 23:53–23:57 UTC), PK Spot experienced a search outage and degraded website performance during a distributed scraping incident. Some website requests continued to succeed; others were slow or failed.
This is the identified incident window. The precise start and end of search unavailability are still being established.
What happened
Our edge logs show approximately 3,550 requests from 1,219 distinct IP addresses during the investigated window. Requests systematically traversed spot pages across all six supported languages, with additional requests to community and event pages. This pattern was consistent with sitemap enumeration.
The traffic used rotating browser user-agent strings, including identities claiming older macOS, Windows, Android and iOS versions. Many requests also supplied a Google Search referrer. These headers did not establish that the requests came from Google; only a handful were identified as genuine Googlebot traffic.
Most source addresses were in AWS-hosted ranges, including addresses beginning with 98.80., 54.91., 13.221., 18.207. and 44.221.. These are representative address prefixes, not attribution to an operator. We have not identified who conducted the scraping or established whether service disruption was their intention.
Why it affected availability
The scraping exposed a bug in our server-side rendering (SSR).
When rendering map-related pages, a consent-handling subscription unintentionally triggered a download of the entire published community list from our search service—even though that data was not needed for the server-rendered page. Repeated page requests therefore generated excessive search traffic from our own backend.
This overwhelmed the search cluster’s resources. As search requests stalled, the absence of an explicit application-level timeout left website renders waiting and occupied backend capacity. Incoming website requests were then delayed or rejected.
The scraping was the trigger, but the amplification bug and insufficient timeout protection were weaknesses in our application.
Impact
- Search: unavailable during the incident.
- Website: degraded performance and intermittent failures, rather than a complete outage.
- Backend: capacity exhaustion produced 1,069 HTTP 429 responses and 16 HTTP 503 responses in the investigated request window.
- Performance: some rendering requests took tens of seconds, with recorded latency reaching approximately 79 seconds.
Current status and remediation
Service has recovered, and we are currently fixing the SSR bug. A fix has been prepared but is not yet being presented as deployed and verified.
Our immediate priorities are to prevent unnecessary community-list downloads during server rendering, add bounded search-request timeouts with graceful failure handling, and verify that concurrent page requests cannot recreate this amplification.
We are also reviewing protections against excessive automated traffic. Those protections complement—not replace—fixing the application behavior that made this incident possible.
We’re sorry for the disruption. We will update this report once the fix has been deployed and validated.
Affected services
Created
Oct 29, 2026 at 11:53pm UTC
Start of the scraping incident was around 00:51 CET.
The issue was first detected around 00:53 CET when our search was timing out.
Affected services