Reconciliation loop i Docs example ?

[Kerwood]

I'm learning kube.rs and I'm reading the docs for how to make a controller.
I was reading this section with an example of reconciliation https://kube.rs/controllers/reconciler/#in-depth-solution

What I'm getting from it is that the example is creating a Controller for a custom resource named PodManager, which will create a new Pod based on some information from the PodManager resource.

The code for the reconcile function is below.

let pod_data = create_owned_pod(&obj);
let serverside = PatchParams::apply("mycontroller");
let pod = pods.patch(pod.name_any(), serverside, Patch::Apply(pod_data)).await?

// update status object with the creation_timestamp of the owned Pod
let status = json!({
    "status": PodManagerStatus { pod_created: pod.meta().creation_timestamp }
});
api.patch_status(&obj.name_any(), &PatchParams::default(), &Patch::Merge(&status))
    .await?;

As you can see, the last step is to update the PodManager status field with pod_created: <timestamp>. But if the controller is created to watch PodManager resources, updating the status field will trigger a new reconciliation, a new timestamp will be set on the PodManager resource and the loop continues.

Am I missing something ?

[nightkr]

The creation_timestamp won't be updated when modifying a Pod (or other object) that already exists, only when creating a new object.

[Kerwood]

Every time that code runs, a new timestamp will be updated on the PodManager resource, right ?

[Kerwood]

ooh creation_timestamp.. Got it.
But the reconcile will run two times right ? When the PodManager gets the timestamp update the first time, the controller will run again, but this time the patch_status() will not update anything because the creation_timestamp has not changed.
Am I understanding it right ?

[nightkr]

Yes. Eventually the reconciliations should reach a stable state where no more modifications will happen and the loop stops.

[Kerwood]

Ok thanks.