Requests to the Triton Inference Endpoint in Modal entered an unstable event loop with 5xx request failures. We spun down the container and launched several more restoring the appropriate number of pods so the system is stable. The system is operational again.
Resolved
Requests to the Triton Inference Endpoint in Modal entered an unstable event loop with 5xx request failures. We spun down the container and launched several more restoring the appropriate number of pods so the system is stable. The system is operational again.
Investigating
API requests are 5xx’ing because containers are failing to spin up. Vendor is undergoing service degradation.