The incident is now resolved and the system is full operational.

The incident is now resolved and the system is full operational.

Between August 26 00:31 to 01:13 UTC, a subset of customers using Inngest Connect experienced delayed and failed function runs due to an internal networking issue that disrupted connectivity between Connect services. We reverted the change and restored service.

Although Connect buffers worker responses, stalled internal API calls delayed acknowledgements and subsequent replies. In some cases, request forwarding failed even after a buffered response had been received.

We fixed the root cause, improved Connect’s failure handling, and enhanced our internal observability so we can detect and respond more quickly to similar problems in the future.

Affected customers should review failed Connect runs during the incident window and replay them after confirming that doing so is safe and idempotent.

Between August 24 at 23:38 UTC and August 25 at 00:08 UTC, a subset of customers using Inngest Connect experienced delayed and failed function runs due to an internal networking issue that disrupted connectivity between Connect services. We reverted the change and restored service.

Although Connect buffers worker responses, stalled internal API calls delayed acknowledgements and subsequent replies. In some cases, request forwarding failed even after a buffered response had been received.

We fixed the root cause, improved Connect’s failure handling, and enhanced our internal observability so we can detect and respond more quickly to similar problems in the future.

Affected customers should review failed Connect runs during the incident window and replay them after confirming that doing so is safe and idempotent.

Run history delay has caught up and the incident is now resolved.

The incident is now resolved and the system is full operational.

Execution latencies are back to normal across all customers, and the incident is now resolved.

We observed temporary function execution delays for a period of 15 minutes as we performed emergency maintenance to manage disk capacity to several nodes.

The incident is now resolved and the system is full operational.

Delays began around 14:03 UTC caused by an issue publishing to a new Kafka topic. Recovery began around 14:40 UTC as the team was able to isolate this topic. The consuming services which schedules new function runs was scaled out and began to consume the backlog. No events were dropped during this incident, all functions were scheduled, but with delays during this time window. No action is needed for manual recovery.

Metrics in the UI are now restored. Users may notice metrics data continue to be unavailable during the time of the incident, from the timeframe 7/16/26 15:45 UTC until approximately 7/16/26 21:07 UTC

Throughput and latency have recovered and the system should be fully operational again.

The core issue is related to the part of the system that persists run state. Some operations experienced and increase in errors and retries, causing a backlog in some parts of the system and some failed operations like checkpoints or signals. We are continuing a thorough post-mortem investigation to prevent reoccurrance.

The incident is now resolved and the system is full operational.

The incident is now resolved and the system is full operational.

The incident is now resolved and the system is full operational.