Intermittent Content Search Failures (Jan 2026)

#Overview

Capable apps (Diagrams, Approval, Calendars, Markdown) are impacted by intermittent Confluence Search API failures, causing unreliable content/user searches and Hystrix/500 errors. Atlassian acknowledges this as P1 but has not posted publicly or provided a timeline for resolution.

#Incident Description

The /wiki/rest/api/search endpoints fail sporadically with 400 BadRequestException: "CQL parsed but search manager unable to execute... Hystrix circuit short-circuited and is OPEN"

Spikes occur ~every 6 hours (45-60 min duration, e.g., 12-13 UTC), affecting all sites. Direct API calls succeed, but calls from Marketplace apps fail, rendering features unusable (e.g., user selection in Approval, search results for diagrams or pages etc.).

#How do I know if I’m affected?

Any features that depend on using the Confluence Content Search API failed sporadically:

• Browsing diagrams from pages

• Viewing templates

• Searching for content using the search feature

• etc.

#Workaround

We've thoroughly investigated alternative APIs to bypass the failing CQL search, but none fully replicate dynamic content/user querying in Forge apps.

The only workaround for this Confluence Search API issue is retrying the operation after a 5-10 minute delay, aligning with observed 45-60 minute spike cycles.

#Incident Timeline