I have a small application which I have build into a docker container, run in K8S and read from a postgresQL DB with Npgsql.

I hit it with 10K req / minute on a scale of 4 pods, so in best load balacing should be 2.5k rq / minute / pod.

On each pod I have a connection pool limit of 100, which is nearly full used here. I exptected to see 1 connection per pod, because it will share the pool (connection string / database is the same) but in fact I see a spike of 200-250 connections total.

Is that a normal behaviour?