Upgrading flagger from 1.40.0 to 1.41.1 causes the canaries to be rollout whithout changes in the deployment

[ealogar]

Describe the bug

In kubernetes 1.32.3 when upgradding flagger version to 1.41.1 all the canaries in the cluster triggers a new rollout without any changes in the deployment that is being targetted.

{"level":"info","ts":"2025-04-15T17:17:58.828Z","caller":"controller/events.go:33","msg":"New revision detected! Scaling upcanary1","canary":"canary1"} {"level":"info","ts":"2025-04-15T17:17:58.830Z","caller":"controller/events.go:33","msg":"New revision detected! Scaling up canary2","canary":"canary2"} ... The same for all the 10 canaries I'm testing

To Reproduce

A k8s cluster in 1.32.3 and some canaries in it update flagger

Expected behavior

Without changes in the deployment no rollout should happen.

Additional context

• Flagger version: 1.41.1
• Kubernetes version: 1.32.3
• Service Mesh provider:
• Ingress provider: nginx

Similar to https://github.com/fluxcd/flagger/issues/1673

[ealogar]

I can reproduce it with kubernetes 1.31.5 as well

[ccystephenclinton]

Seeing this too with version 1.41.1 on Kubernetes version 1.31.4 in Amazon EKS

We have a suspicion it's being caused by .spec.service.headless https://github.com/fluxcd/flagger/blob/main/CHANGELOG.md#1410

[vkoshkarovroku]

We're seeing the same issue. Still not sure what in Flagger causes that...