I have a kubernetes cluster, created and managed from stackpoint.io.

The cluster is made of 1 master and 2 nodes, all running on coreOs.

I created 5 deployments with 1 replicas each and 5 services pointing to these deployments.

5 pods are created from these deployments.

My problem is : between 24 to 36 hours after the pods are created, they are automatically deleted and recreated.

because these different apps rely on each other to work, it takes about 3-5 minutes of downtime before everything works correctly again.

I assume there is some configuration I don't know about that control this behavior.

I tested using containers with latest tag as well as not latest. I tested changing the pullStrategie from Always to IfNotPresent. I tested using a ubuntu based cluster instead of CoreOs. In every configuration the pods are deleted and recreated