Elevated error rate on gpuserver.io

时间线

1. 已解决2026年7月4日 11:12 UTC

   Resolved at 11:12 UTC. Duration: 12 minutes. Order requests submitted during the window were either completed normally or shown an error immediately; none was silently lost.

2. 已确定原因2026年7月4日 11:09 UTC

   A deployment at 11:00 UTC exhausted the application worker pool under an unexpected request pattern. We have rolled the deployment back.

3. 调查中2026年7月4日 11:06 UTC

   About 32% of requests to gpuserver.io are returning a 502 since 11:00 UTC. The configurator is affected. Machines, payments and provisioning are not.

事故复盘会在解决后 5 天内发布于此。