Benjamin Satzger a49d018713 Avoid creating short-lived pages for inference endpoints ...

The load_balancers_vms table includes state and state_counter columns
for each VM associated with a load balancer. To make this relationship
explicit for inference endpoint replicas, we establish a
load_balancers_vm relationship.

This relationship serves two key purposes:
  1. To determine the availability of a replica.
  2. To ensure that a page is created only if a replica has been down
     for a duration exceeding the health_check_down_threshold.

2024-12-04 17:28:07 +01:00

1.1 KiB

Raw Permalink Blame History

View Raw