Pires

The problem is not the master but the worker nodes. However, it would be super cool to have it working fully for Linux and Windows.

I think it's a great idea. I just won't be able to work on that, for the time being.

This is definitely something we're looking forward to have merged, as well.

Unless I'm missing something here, the configuration is passed by the user.

The issue title is misleading. In this scenario, the **operator exits only when it loses leadership**, either because: - another instance of the operator won the leader lease [1], or...

@wallyqs want to pick it up? :)

Cluster-scoped operator instance(s) may overlap with namespace-scoped operator instance(s) because they'll rely on different namespaces for doing leader-election. This means you'll have two or more instances of the operator managing...

> I don't understand why we force the use of the `nats-io` namespace here. Using a fixed, well-known namespace - as opposed to allowing for it to be specified using...

> Many operators are already doing it already (never checked how it's done right now, sorry). Can you provide those many examples (links to the code, please)? I'd like to...

That's exactly what this operator uses (client-go leaderelection)! We simply use an `EndpointLock`, while `cert-manager` uses a `ConfigMapLock`. However, `cert-manager` implementation does allow for the configuration of the namespace to...