Summary:
A critical backend infrastructure deadlock on a private deployment is generating continuous, unauthorized billing charges despite the deployment being Disabled and max_instances explicitly set to 0.
Account & Environment Details:
- Username: aurorarecmgmt-cloud
- Affected Resource: Private Deployment (huggy-mochi-generator)
- Current Status: Disabled, max_instances: 0, API Token Revoked.
Bug Description & Evidence:
- Infrastructure Deadlock: The deployment metric panel shows SETUP TIME: 100.0%, ACTIVE TIME: 0.0%, and Queue length: 1. The backend hardware pod is stuck in an infinite initialization/cold-start loop and refuses to terminate.
- Uncontrolled Charges: The outstanding balance jumped from ~$9.00 USD (when the initial support ticket was submitted 18 hours ago) to over $91.46 USD in 18+ hours while completely inactive, with zero user predictions being processed.
- Deletion Blocked: Attempting to delete the deployment fails due to lingering stuck background instances.
No Support Response: First email was sent to support@replicate.com 18 hours ago when the balance was ~$9.00 USD. No response has been received while costs continue to accumulate rapidly.
Expected Behavior:
Setting max_instances to 0 and disabling the deployment should immediately terminate all underlying hardware pods and stop all billing ticks.
Actual Behavior:
The backend pod remains locked in the setup phase indefinitely, continuously accruing charges on a disabled deployment.
Screenshots Attached:
- Metric dashboard showing SETUP TIME: 100% / ACTIVE TIME: 0%
- Billing dashboard showing unpaid balance surging to $91.46 USD
- Deployment settings showing Disabled and max_instances: 0
Requested Resolution:
- Force Terminate: Immediately kill all background pods/instances tied to account aurorarecmgmt-cloud.
- Billing Adjustment: Reset/refund the erroneous charges accumulated during this platform setup deadlock back to $0.00.
(Escalating here for engineering visibility due to zero response on the 18-hour-old support ticket.)
Summary:
A critical backend infrastructure deadlock on a private deployment is generating continuous, unauthorized billing charges despite the deployment being Disabled and max_instances explicitly set to 0.
Account & Environment Details:
Bug Description & Evidence:
No Support Response: First email was sent to support@replicate.com 18 hours ago when the balance was ~$9.00 USD. No response has been received while costs continue to accumulate rapidly.
Expected Behavior:
Setting max_instances to 0 and disabling the deployment should immediately terminate all underlying hardware pods and stop all billing ticks.
Actual Behavior:
The backend pod remains locked in the setup phase indefinitely, continuously accruing charges on a disabled deployment.
Screenshots Attached:
Requested Resolution:
(Escalating here for engineering visibility due to zero response on the 18-hour-old support ticket.)