everyy time we push a new build to the production cluster, the load balancer starts dropping requests. it seems like the
health check intervals are too long to catch the momentary service downtime. i noticed that the error logs specifically show
504 gateway timeouts right when the containers restart. trying to fix this by manually adjusting the
max_unavailable
setting in the deployment manifest helped a bit.
>the traffic just keeps hitting dead nodesit is pretty frustrating bc it makes every release feel like a high-stakes gamble instead of a routine task.