The NetworkManager logic was implemented directly into the `HaClusterTopology`, which wasn't directly its concern and prevented us from being able to reuse that NetworkManaager implementations in the future for a different Topology.
* Extract a `NetworkManager` trait
* Implement a `OpenShiftNmStateNetworkManager` for `NetworkManager`
* Dynamically instantiate the NetworkManager in the Topology to delegate calls to it
Reviewed-on: #183
Reviewed-by: johnride <jg@nationtech.io>
In order to query the current network state `NodeNetworkState` and to apply a `NodeNetworkConfigurationPolicy` for a given node, we first needed to find its hostname. As all we had was the UUID of a node.
We had different options available (e.g. updating the Harmony Inventory Agent to retrieve it, store it in the OKD installation pipeline on assignation, etc.). But for the sake of simplicity and for better flexibility (e.g. being able to run this score on a cluster that wasn't setup with Harmony), the `hostname` was retrieved directly in the cluster by running the equivalent of `kubectl get nodes -o yaml` and matching the nodes with the system UUID.
### Other changes
* Find the next available bond id for a node
* Apply a network config policy for a node (configuring a bond in our case)
* Adjust the CRDs for NMState
Note: to see a quick demo, watch the recording in #183
Reviewed-on: #182
Reviewed-by: johnride <jg@nationtech.io>
With this work, no need to initialize a full HAClusterTopology to run
opnsense scores.
Also added an example showing how to use it and perform basic
operations.
Made a video out of it, might publish it at some point!
## Description
* Replace the CatalogSource approach to install the OperatorHub.io catalog by a more simple & straightforward way to install NMState
* Improve logging
* Add report summarizing the host network configuration that was applied (which host, bonds, port-channels)
* Fix command to find next available port channel id
## Extra info
Using the `apply_url` approach to install the NMState operator isn't the best approach: it's harder to maintain and upgrade. But it helps us achieve waht we wanted for now: install the NMState Operator to configure bonds on a host.
The preferred approach, installing an operator from the OperatorHub.io catalog, didn't work for now. We had a timeout error with DeadlineExceeded probably caused by an insufficient CPU/Memory allocation to query such a big catalog, even though we tweaked the RAM allocation (we couldn't find a way to do it for CPU).
Spent too much time on this so we stopped these efforts for now. It would be good to get back to it when we need to install something else from a custom catalog.
Reviewed-on: #175
k8s returns None rather than zero when checking deployment for replicas
exec_app requires commands 's' and '-c' to run correctly
Reviewed-on: #120
Co-authored-by: Willem <wrolleman@nationtech.io>
Co-committed-by: Willem <wrolleman@nationtech.io>
The previous implementation blindly added HAProxy components without checking for existing configurations on the same port, which caused duplicate entries and errors when a service was updated.
This commit refactors the logic to a robust "remove-then-add" strategy. The configure_service method now finds and removes any existing frontend and its dependent components (backend, servers, health check) before adding the new, complete service definition.
This change makes the process fully idempotent, preventing configuration drift and ensuring a predictable state.
Co-authored-by: Ian Letourneau <letourneau.ian@gmail.com>
Reviewed-on: #129
