[Bug]: excessive sleep in patroni_core_controller.go

[vlsi]

Describe the bug

See

pgskipper-operator/operator/controllers/patroni_core_controller.go

Lines 196 to 197 in 6dbb712

	pr.logger.Info("Reconcile will be started...")
	time.Sleep(30 * time.Second)

The sleep is not documented, and it looks like it spends time without explanation.
At least the code should document the sleep or it should be replaced with a conditional wait.

To Reproduce

See

pgskipper-operator/operator/controllers/patroni_core_controller.go

Lines 196 to 197 in 6dbb712

	pr.logger.Info("Reconcile will be started...")
	time.Sleep(30 * time.Second)

Version

No response

Logs

No response

Additional information

No response

[Tvion]

This sleep is required for operator in the case of upgrade. A delay may occur if the CR has been updated by Helm, but the operator has not yet been updated, and the old operator has begun reconciling with the new CR.

[vlsi]

This sleep is required for operator in the case of upgrade

Why is it a seep(30 sec) rather than "wait for a specific condition"?

[Tvion]

What condition do you propose?

[vlsi]

The code assumes that "something will complete within 30 sec". So it should check the status of that "something", shouldn't it?

[Tvion]

Yes, but this wait operation is a part of condition. Condition here may be described as: "If current pod won't be deleted within 30 sec, then continue". It's probably should be some improvement related to CR versioning, so operator won't process CR with versions that belong to different releases. It looks like a feature, but not bugfix.

However we are planning to migrate to cluster wide operator, so current issue won't be actual.