Beginning at 23:17 UTC on September 24, 2025 and continuing until 00:39 UTC on September 25, 2025, impacted customers experienced elevated H99 errors on network runtime in US & EU. This issue is now resolved.
Heroku engineers are investigating a decline in successful API responses from the gpt-oss-120b AI model in US region.
The upstream DNS issue impacting Heroku has been resolved. Customers should no longer see errors, though some may continue to experience degraded performance as services stabilize. We are actively monitoring the situation to ensure full reliability and will provide further updates as needed.
Configuring new connections within the Heroku Connect add-on through the dashboard is functioning but requires a workaround. Currently, users must cancel at step 2 of the connection configuration and refresh to see the connection status be successful. Our engineers are working to resolve the issue.
Engineers are deploying a kernel update to Common Runtime fleets to fix an issue causing frequent dyno relocations.