instructions.md

Overview

We are trying to replicate a migration scenario where src cluster hpa may be disabled/differ from target as we are scaling to 0.

Expected results

We expect the src and target deployments replica count to be able to be controlled independently, by only their respective hpa.

Actual results

Target cluster replica count is being continuously updated by karmada by the src deployment/hpas replica count.

steps to reproduce

create deployment, hpa

ka source apply -f deployment.yaml

ka source apply -f hpa.yaml

note that this hpa has scaling disabled and minReplicas = maxReplicas. So all this hpa should be doing is changing the replica count to the minReplicas value (2 in this case)

propagate to target cluster

ka source apply -f pp.yaml

observe matching state in both clusters

ka source -n retain-test get deployment

ka member1 -n retain-test get deployment

apply retain label to target cluster. replica count should still match in both clusters, should be able to observe label applied

ka source apply -f retain-label-op.yaml

update target cluster hpa

ka source apply -f hpa.yaml

note we are updating minReplicas/maxReplicas to 4, and hpa is still disabled. this should just scale the target deployment to 4

observe deployment state

ka source -n retain-test get deployment -w

ka member1 -n retain-test get deployment -w

we see that replica count is spamming between 2 and 4, when it should be constant at 4 in target cluster