controller: reject adding ifaces to host networked pods #89
Conversation
The controller unit tests were refactored a bit - creating the k8s client and the dummy controller - in a `JustBeforeEach` block to allow each `When` / `Context` block to define the pod previously. This way, there is a lot less duplicated code. Signed-off-by: Miguel Duarte Barroso <[email protected]>
| "rejecting to add interfaces for host networked pod: %s", | ||
| annotations.NamespacedName(newPod.GetNamespace(), newPod.GetName()), | ||
| ) | ||
| pnc.Eventf(newPod, corev1.EventTypeWarning, "InterfaceAddRejected", rejectInterfaceAddEventFormat(newPod)) |
There was a problem hiding this comment.
I'm assuming we want to both throw an event an log ... Maybe that's too much.
Any thoughts @kmabda ?
There was a problem hiding this comment.
I think it's a good idea to throw an event here to give a feedback to the user
There was a problem hiding this comment.
One thing, since in https://github.com/k8snetworkplumbingwg/multus-dynamic-networks-controller/pull/90/files#diff-449e8d1bd46b9c978e208dde912a1d7a76d17ce553067f8d0a67a27f17338d26R152 we need to filter out hostnetwork pods as well but we can't tel before hand if an interface got added by the user or not until we process the request we could move this logic from handlePodUpdate to processNextWorkItem function.
Pro same treatment logic would be in one place i.e. we have a consistency w.r.t. filtering out unwanted pods and how we log/throw out an event to the user in case an interface got added for these pods in one place
Con a little more overhead by adding an item to the workqueue to processing it than throwing away
If you go ahead with these changes I can rebase my pull request once this one get merged.
What this PR does / why we need it:
This PR rejects adding / removing interfaces to/from host networked pods.
The controller unit tests were refactored a bit - creating the k8s client and the dummy controller - in a
JustBeforeEachblock to allow eachWhen/Contextblock to define the pod previously. This way, there is a lot less duplicated code.Which issue(s) this PR fixes (optional, in
fixes #<issue number>(, fixes #<issue_number>, ...)format, will close the issue(s) when PR gets merged):Fixes #88
Special notes for your reviewer (optional):