LiveKit Outage History

LiveKit experienced elevated participant connection latency and errors in the US East region due to database connection timeouts, affecting all services in that region for 5.4 hours. The issue was resolved by draining traffic from US East and routing it to other regions, which eliminated the connection errors. Service returned to normal operation with traffic continuing to route through alternative regions while US East was being restored.

LiveKit Cloud Dashboard experienced a partial data outage affecting Agent Observability access for some sessions and causing ended sessions to incorrectly display as active in the 24-hour filter. The issue was identified as partial data unavailability, and a fix was implemented to prevent further data loss while missing data was backfilled. The incident was resolved after 2.2 hours with all missing data successfully restored.

LiveKit experienced network instability in Frankfurt starting at 18:15 UTC that impacted cross-region traffic. Traffic was redirected away from Frankfurt by 19:55 UTC, causing participants who originally connected through that region to experience slightly increased latency. The Frankfurt cluster was re-enabled after 20.2 hours and monitored for 16 additional hours before the incident was fully resolved.

LiveKit experienced intermittent connectivity disruptions affecting a subset of sessions in US regions, particularly US East, due to internal network degradation between clusters. The disruptions caused brief interruptions (less than 5 minutes) to media relay across multiple sessions at various times between May 12-15, 2026. The incident was resolved after 9.9 hours when no further connectivity issues were observed, with the team continuing to investigate why their automatic reroute mechanism failed to activate during these isolated instances.

LiveKit's Global SIP service experienced a major outage where outbound TLS calls were failing for approximately 3.3 hours. The issue was initially thought to be Twilio-specific but was later determined to affect all outbound TLS calls regardless of provider. A fix was implemented and the incident was resolved after monitoring confirmed no further errors.

LiveKit Cloud Agents in the US East region experienced deployment failures and degraded performance for 5.7 hours due to an etcd service outage that caused Kubernetes API server unresponsiveness. The incident initially affected only new agent deployments but later expanded to impact existing deployments as well. The service was fully restored after the Kubernetes infrastructure was repaired with assistance from the data center provider.

LiveKit experienced SIP participant timeouts in Singapore and increased signaling connection errors primarily affecting India, with potential impact on connection latency and inbound SIP calls across multiple regions including Australia, US Central, and South Africa. The issue affected a very small minority of requests and was resolved through fine-tuning adjustments to improve stack performance. The incident lasted 116.5 hours with minimal user reports throughout the duration.

LiveKit experienced increased latencies in RoomService APIs (CreateRoom, DeleteRoom, and UpdateRoomMetadata) that initially affected the US West region but later expanded to other regions, caused by issues with a specific table in their distributed database. During the 4.8-hour incident, users experienced slow API responses and a temporary 10% failure rate when the fix was being applied. The issue was fully resolved after implementing a workaround and working with the database vendor.

The LiveKit Cloud Dashboard (cloud.livekit.io) experienced a major outage that prevented users from accessing the service. The incident lasted 13 minutes before a fix was implemented and normal access was restored.

LiveKit experienced degraded connectivity issues affecting Real Time Communication and TURN services in the EU (Frankfurt) region for 2.5 hours. The service team identified the root cause and implemented a fix to restore normal connectivity. The incident was resolved after a monitoring period to confirm service stability.