Resolved
This incident is resolved. Parse latency has been normal since 8:35 PM PDT on September 27.
Monitoring
The GPU clusters are recovered, and Parse latency is back to normal. We are monitoring the results.
Identified
Parse latency is degraded because of an issue with our GPU clusters. Our engineers are working on a fix. Only the /parse endpoint is affected.
Investigating
Parse requests are degraded. Only the /parse endpoint is affected. Extract, Split, and Edit are operating normally. We are investigating.
