Forge invocation outage, January 2026

Updated
View as text
Open in Confluence
Open in ChatGPT Open in Claude

Severity

P1

Duration

2 hours 40 minutes

Atlassian ticket

ECOHELP-109057

#What happened

On 28 January 2026 at approximately 16:20 UTC, the calls our apps make to the Atlassian Forge platform began failing and timing out. This affected every Capable app built on Forge: Diagrams, Approval, Calendar and Markdown among them.

The cause was not in our code. The Forge invocation layer was failing platform-wide because of errors in Atlassian's authentication system, and other Marketplace vendors reported identical symptoms within the hour.

#Timeline (UTC)

Time

Event

16:20

We detected the failures in production

16:29

First report raised with Atlassian

17:00

Other Marketplace partners confirmed the same symptoms

17:34

Atlassian publicly acknowledged the incident

#How you would have known

A macro placeholder in a Confluence page showing a spinner above the words Loading diagram.
Macros behaved the same way, holding their loading state instead of drawing anything.
The byline row of a Confluence page titled All macros, showing byline items from several apps, with the Capable item still displaying a loading spinner and outlined in red.
The byline item spins instead of resolving, which is what most people noticed first.

  • Byline items taking a long time to load, or not loading at all

  • Macros taking a long time to load, or not loading at all

  • Blank app panels, or an error message where a macro should be

#Customer data

#Resolution

Nothing was required from customers. Service returned when Atlassian restored the Forge invocation layer.

Written up, kept, and linked from the docs.