LaunchDarkly Outage History

Experimentation results processing and data warehouse export was caught up as of 12:28 pm PT.

We've identified elevated error rates with server-side streaming initialization requests in the APAC region between 19:52 and 19:56 PT. Some customers may have experienced initialization timeouts despite SDK retries.

We've caught up on the streaming data export backlog.

We've identified an infrastructure issue that caused some 503 responses in feature flagging streaming client initialization on the US LaunchDarkly instance between 16:45 and 17:05 PT. SDKs should automatically retry initialization in most cases.

We have validated a fix to resolve all remaining experiments with inaccurate sample ratio mismatches (SRMs) or empty experiment results. The vast majority of experiments and experimentation customers were not impacted.

Observability alerts are functioning and delayed alerts have been redelivered.

We've fully caught up on the metrics data backlog.

We're addressing an issue that prevented warehouse data export from 12pm to 6pm PT on April 23 and are working to backfill that data to customer warehouses.

This incident has been resolved.

The issue has been resolved.

Between approximately 9:37 PM and 9:46 PM PT on April 9, 2026, some customers using server-side SDKs in the EU region may have experienced longer than normal initialization times or timeouts when connecting to LaunchDarkly. The issue resolved. No action is required from customers. Server-side SDKs automatically reconnect and recover from transient connectivity issues.

The delayed processing of Session Replay and Error event data has been resolved. All queued events have been processed and data ingestion is operating normally. No data was lost during this incident. Thank you for your patience.

Server-side streaming began rejecting new connections across all commercial regions, causing 500/503 errors for customers attempting to establish new SDK streaming connections for a brief period of time. Detailed timelines of the incident: - us-east-1 : 7:53 AM PST to 8:24 AM PST - ap-southeast-1: 7:53 AM PST to 7:55 AM PST - eu-west-1: 7:53 AM PST to 8:35 AM PST The team was quickly able to deploy the mitigation before an incident could be declared in our status pages, this is why we're posting it, retroactively. By 8:35am PST connections were re-established successfully in all commercial regions.

Our team completed the data restoration for previously affected experiment iterations. Complete data is now available for all experiment iterations in the UI.

The issue with Event processing has been resolved. Impacted services have returned to normal operation. Flag Delivery was not impacted. There was no data loss experienced.

We've corrected the issues affecting the Observability "Errors" usage reporting for February and are finalizing reporting corrections for March.

The issue with our application and authentication has been resolved. Performance has remained stable following mitigation.

Ingest performance has remained stable following mitigation. We are no longer observing impact.

We've caught up on the session/error data backlog.

We have fixed the data attribution issues where running experiment iterations were receiving incorrectly attributed data. This incident is now resolved and results for active experiments are now accurate.

This incident has been resolved. Telemetry data was dropped between 5:03 PM and 7:30 PM PT. All systems are now operating normally.

This issue has been resolved. Observability ingestion from AWS CloudWatch Metric Streams and CloudWatch Firehose log export has returned to normal. Data sent during the incident window may not have been ingested, and customers may see gaps in metrics or logs for that period.

Customers are seeing expected behavior when accessing the LaunchDarkly UI and API.