Compare commits

..

7 Commits

Author SHA1 Message Date
4f4510d7ba feat: added steps for bootstrap install worker node
All checks were successful
Run Check Script / check (pull_request) Successful in 1m24s
2025-11-24 14:56:37 -05:00
83c1cc82b6 fix(host_network): remove extra fields from bond config to prevent clashes (#186)
All checks were successful
Run Check Script / check (push) Successful in 1m36s
Compile and package harmony_composer / package_harmony_composer (push) Successful in 8m16s
Also alias `port` to support both `port` and `ports` as per the nmstate spec.

Reviewed-on: #186
2025-11-11 14:12:56 +00:00
66d346a10c fix(host_network): skip configuration for host with only 1 interface/port (#185)
All checks were successful
Run Check Script / check (push) Successful in 1m11s
Compile and package harmony_composer / package_harmony_composer (push) Successful in 8m11s
Reviewed-on: #185
Reviewed-by: johnride <jg@nationtech.io>
2025-11-06 00:07:20 +00:00
06a004a65d refactor(host_network): extract NetworkManager as a reusable component (#183)
Some checks failed
Run Check Script / check (push) Successful in 1m12s
Compile and package harmony_composer / package_harmony_composer (push) Has been cancelled
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>
2025-11-06 00:02:52 +00:00
9d4e6acac0 fix(host_network): retrieve proper hostname and next available bond id (#182)
Some checks failed
Run Check Script / check (push) Successful in 1m9s
Compile and package harmony_composer / package_harmony_composer (push) Failing after 2m24s
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>
2025-11-05 23:38:24 +00:00
4ff57062ae Merge pull request 'feat(kube): Convert kube_openapi Resource to DynamicObject' (#180) from feat/kube_convert_dynamic_resource into master
Some checks failed
Run Check Script / check (push) Successful in 1m19s
Compile and package harmony_composer / package_harmony_composer (push) Failing after 2m23s
Reviewed-on: #180
Reviewed-by: Ian Letourneau <ian@noma.to>
2025-11-05 21:48:32 +00:00
95cfc03518 feat(kube): Utility function to convert kube_openapi Resource to DynamicObject. This will allow initializing resources strongly typed and then bundle various types into a list of DynamicObject
All checks were successful
Run Check Script / check (pull_request) Successful in 1m18s
2025-10-29 17:24:35 -04:00
8 changed files with 375 additions and 21 deletions

View File

@@ -1,6 +1,7 @@
use async_trait::async_trait;
use harmony_macros::ip;
use harmony_types::{
id::Id,
net::{MacAddress, Url},
switch::PortLocation,
};

182
harmony/src/infra/kube.rs Normal file
View File

@@ -0,0 +1,182 @@
use k8s_openapi::Resource as K8sResource;
use kube::api::{ApiResource, DynamicObject, GroupVersionKind};
use kube::core::TypeMeta;
use serde::Serialize;
use serde::de::DeserializeOwned;
use serde_json::Value;
/// Convert a typed Kubernetes resource `K` into a `DynamicObject`.
///
/// Requirements:
/// - `K` must be a k8s_openapi resource (provides static GVK via `Resource`).
/// - `K` must have standard Kubernetes shape (metadata + payload fields).
///
/// Notes:
/// - We set `types` (apiVersion/kind) and copy `metadata`.
/// - We place the remaining top-level fields into `obj.data` as JSON.
/// - Scope is not encoded on the object itself; you still need the corresponding
/// `DynamicResource` (derived from K::group/version/kind) when constructing an Api.
///
/// Example usage:
/// let dyn_obj = kube_resource_to_dynamic(secret)?;
/// let api: Api<DynamicObject> = Api::namespaced_with(client, "ns", &dr);
/// api.patch(&dyn_obj.name_any(), &PatchParams::apply("mgr"), &Patch::Apply(dyn_obj)).await?;
pub fn kube_resource_to_dynamic<K>(res: &K) -> Result<DynamicObject, String>
where
K: K8sResource + Serialize + DeserializeOwned,
{
// Serialize the typed resource to JSON so we can split metadata and payload
let mut v = serde_json::to_value(res).map_err(|e| format!("Failed to serialize : {e}"))?;
let obj = v
.as_object_mut()
.ok_or_else(|| "expected object JSON".to_string())?;
// Extract and parse metadata into kube::core::ObjectMeta
let metadata_value = obj
.remove("metadata")
.ok_or_else(|| "missing metadata".to_string())?;
let metadata: kube::core::ObjectMeta = serde_json::from_value(metadata_value)
.map_err(|e| format!("Failed to deserialize : {e}"))?;
// Name is required for DynamicObject::new; prefer metadata.name
let name = metadata
.name
.clone()
.ok_or_else(|| "metadata.name is required".to_string())?;
// Remaining fields (spec/status/data/etc.) become the dynamic payload
let payload = Value::Object(obj.clone());
// Construct the DynamicObject
let mut dyn_obj = DynamicObject::new(
&name,
&ApiResource::from_gvk(&GroupVersionKind::gvk(K::GROUP, K::VERSION, K::KIND)),
);
dyn_obj.types = Some(TypeMeta {
api_version: api_version_for::<K>(),
kind: K::KIND.into(),
});
// Preserve namespace/labels/annotations/etc.
dyn_obj.metadata = metadata;
// Attach payload
dyn_obj.data = payload;
Ok(dyn_obj)
}
/// Helper: compute apiVersion string ("group/version" or "v1" for core).
fn api_version_for<K>() -> String
where
K: K8sResource,
{
let group = K::GROUP;
let version = K::VERSION;
if group.is_empty() {
version.to_string() // core/v1 => "v1"
} else {
format!("{}/{}", group, version)
}
}
#[cfg(test)]
mod test {
use super::*;
use k8s_openapi::api::{
apps::v1::{Deployment, DeploymentSpec},
core::v1::{PodTemplateSpec, Secret},
};
use kube::api::ObjectMeta;
use pretty_assertions::assert_eq;
#[test]
fn secret_to_dynamic_roundtrip() {
// Create a sample Secret resource
let mut secret = Secret {
metadata: ObjectMeta {
name: Some("my-secret".to_string()),
..Default::default()
},
type_: Some("kubernetes.io/service-account-token".to_string()),
..Default::default()
};
// Convert to DynamicResource
let dynamic: DynamicObject =
kube_resource_to_dynamic(&secret).expect("Failed to convert Secret to DynamicResource");
// Serialize both the original and dynamic resources to Value
let original_value = serde_json::to_value(&secret).expect("Failed to serialize Secret");
let dynamic_value =
serde_json::to_value(&dynamic).expect("Failed to serialize DynamicResource");
// Assert that they are identical
assert_eq!(original_value, dynamic_value);
secret.metadata.namespace = Some("false".to_string());
let modified_value = serde_json::to_value(&secret).expect("Failed to serialize Secret");
assert_ne!(modified_value, dynamic_value);
}
#[test]
fn deployment_to_dynamic_roundtrip() {
// Create a sample Deployment with nested structures
let mut deployment = Deployment {
metadata: ObjectMeta {
name: Some("my-deployment".to_string()),
labels: Some({
let mut map = std::collections::BTreeMap::new();
map.insert("app".to_string(), "nginx".to_string());
map
}),
..Default::default()
},
spec: Some(DeploymentSpec {
replicas: Some(3),
selector: Default::default(),
template: PodTemplateSpec {
metadata: Some(ObjectMeta {
labels: Some({
let mut map = std::collections::BTreeMap::new();
map.insert("app".to_string(), "nginx".to_string());
map
}),
..Default::default()
}),
spec: Some(Default::default()), // PodSpec with empty containers for simplicity
},
..Default::default()
}),
..Default::default()
};
let dynamic = kube_resource_to_dynamic(&deployment).expect("Failed to convert Deployment");
let original_value = serde_json::to_value(&deployment).unwrap();
let dynamic_value = serde_json::to_value(&dynamic).unwrap();
assert_eq!(original_value, dynamic_value);
assert_eq!(
dynamic.data.get("spec").unwrap().get("replicas").unwrap(),
3
);
assert_eq!(
dynamic
.data
.get("spec")
.unwrap()
.get("template")
.unwrap()
.get("metadata")
.unwrap()
.get("labels")
.unwrap()
.get("app")
.unwrap()
.as_str()
.unwrap(),
"nginx".to_string()
);
}
}

View File

@@ -3,6 +3,7 @@ pub mod executors;
pub mod hp_ilo;
pub mod intel_amt;
pub mod inventory;
pub mod kube;
pub mod network_manager;
pub mod opnsense;
mod sqlx;

View File

@@ -135,8 +135,6 @@ impl OpenShiftNmStateNetworkManager {
description: Some(format!("Member of bond {bond_name}")),
r#type: nmstate::InterfaceType::Ethernet,
state: "up".to_string(),
mtu: Some(switch_port.interface.mtu),
mac_address: Some(switch_port.interface.mac_address.to_string()),
ipv4: Some(nmstate::IpStackSpec {
enabled: Some(false),
..Default::default()
@@ -162,7 +160,7 @@ impl OpenShiftNmStateNetworkManager {
interfaces.push(nmstate::Interface {
name: bond_name.to_string(),
description: Some(format!("Network bond for host {host}")),
description: Some(format!("HARMONY - Network bond for host {host}")),
r#type: nmstate::InterfaceType::Bond,
state: "up".to_string(),
copy_mac_from,
@@ -241,7 +239,7 @@ impl OpenShiftNmStateNetworkManager {
.and_then(|network_state| network_state.status.current_state.as_ref())
.map_or(&interfaces, |current_state| &current_state.interfaces)
.iter()
.filter(|i| i.r#type == nmstate::InterfaceType::Bond && i.link_aggregation.is_some())
.filter(|i| i.r#type == nmstate::InterfaceType::Bond)
.collect();
let used_ids: HashSet<u32> = existing_bonds

View File

@@ -1,15 +1,21 @@
use async_trait::async_trait;
use derive_new::new;
use harmony_types::id::Id;
use log::info;
use log::{debug, info};
use serde::Serialize;
use crate::{
data::Version,
hardware::PhysicalHost,
infra::inventory::InventoryRepositoryFactory,
interpret::{Interpret, InterpretError, InterpretName, InterpretStatus, Outcome},
inventory::Inventory,
inventory::{HostRole, Inventory},
modules::{
dhcp::DhcpHostBindingScore, http::IPxeMacBootFileScore,
inventory::DiscoverHostForRoleScore, okd::templates::BootstrapIpxeTpl,
},
score::Score,
topology::HAClusterTopology,
topology::{HAClusterTopology, HostBinding},
};
// -------------------------------------------------------------------------------------------------
@@ -52,6 +58,159 @@ impl OKDSetup04WorkersInterpret {
info!("[Workers] Rendering per-MAC PXE for workers and rebooting");
Ok(())
}
/// Ensures that three physical hosts are discovered and available for the ControlPlane role.
/// It will trigger discovery if not enough hosts are found.
async fn get_nodes(
&self,
inventory: &Inventory,
topology: &HAClusterTopology,
) -> Result<Vec<PhysicalHost>, InterpretError> {
const REQUIRED_HOSTS: usize = 2;
let repo = InventoryRepositoryFactory::build().await?;
let mut control_plane_hosts = repo.get_host_for_role(&HostRole::Worker).await?;
while control_plane_hosts.len() < REQUIRED_HOSTS {
info!(
"Discovery of {} control plane hosts in progress, current number {}",
REQUIRED_HOSTS,
control_plane_hosts.len()
);
// This score triggers the discovery agent for a specific role.
DiscoverHostForRoleScore {
role: HostRole::Worker,
}
.interpret(inventory, topology)
.await?;
control_plane_hosts = repo.get_host_for_role(&HostRole::Worker).await?;
}
if control_plane_hosts.len() < REQUIRED_HOSTS {
Err(InterpretError::new(format!(
"OKD Requires at least {} control plane hosts, but only found {}. Cannot proceed.",
REQUIRED_HOSTS,
control_plane_hosts.len()
)))
} else {
// Take exactly the number of required hosts to ensure consistency.
Ok(control_plane_hosts
.into_iter()
.take(REQUIRED_HOSTS)
.collect())
}
}
/// Configures DHCP host bindings for all control plane nodes.
async fn configure_host_binding(
&self,
inventory: &Inventory,
topology: &HAClusterTopology,
nodes: &Vec<PhysicalHost>,
) -> Result<(), InterpretError> {
info!("[Worker] Configuring host bindings for worker nodes.");
// Ensure the topology definition matches the number of physical nodes found.
if topology.control_plane.len() != nodes.len() {
return Err(InterpretError::new(format!(
"Mismatch between logical control plane hosts defined in topology ({}) and physical nodes found ({}).",
topology.control_plane.len(),
nodes.len()
)));
}
// Create a binding for each physical host to its corresponding logical host.
let bindings: Vec<HostBinding> = topology
.control_plane
.iter()
.zip(nodes.iter())
.map(|(logical_host, physical_host)| {
info!(
"Creating binding: Logical Host '{}' -> Physical Host ID '{}'",
logical_host.name, physical_host.id
);
HostBinding {
logical_host: logical_host.clone(),
physical_host: physical_host.clone(),
}
})
.collect();
DhcpHostBindingScore {
host_binding: bindings,
domain: Some(topology.domain_name.clone()),
}
.interpret(inventory, topology)
.await?;
Ok(())
}
/// Renders and deploys a per-MAC iPXE boot file for each control plane node.
async fn configure_ipxe(
&self,
inventory: &Inventory,
topology: &HAClusterTopology,
nodes: &Vec<PhysicalHost>,
) -> Result<(), InterpretError> {
info!("[Worker] Rendering per-MAC iPXE configurations.");
// The iPXE script content is the same for all control plane nodes,
// pointing to the 'master.ign' ignition file.
let content = BootstrapIpxeTpl {
http_ip: &topology.http_server.get_ip().to_string(),
scos_path: "scos",
ignition_http_path: "okd_ignition_files",
installation_device: "/dev/sda", // This might need to be configurable per-host in the future
ignition_file_name: "worker.ign", // Worker nodes use the worker ignition file
}
.to_string();
debug!("[Worker] iPXE content template:\n{content}");
// Create and apply an iPXE boot file for each node.
for node in nodes {
let mac_address = node.get_mac_address();
if mac_address.is_empty() {
return Err(InterpretError::new(format!(
"Physical host with ID '{}' has no MAC addresses defined.",
node.id
)));
}
info!(
"[Worker] Applying iPXE config for node ID '{}' with MACs: {:?}",
node.id, mac_address
);
IPxeMacBootFileScore {
mac_address,
content: content.clone(),
}
.interpret(inventory, topology)
.await?;
}
Ok(())
}
/// Prompts the user to reboot the target control plane nodes.
async fn reboot_targets(&self, nodes: &Vec<PhysicalHost>) -> Result<(), InterpretError> {
let node_ids: Vec<String> = nodes.iter().map(|n| n.id.to_string()).collect();
info!("[Worker] Requesting reboot for control plane nodes: {node_ids:?}",);
let confirmation = inquire::Confirm::new(
&format!("Please reboot the {} worker nodes ({}) to apply their PXE configuration. Press enter when ready.", nodes.len(), node_ids.join(", ")),
)
.prompt()
.map_err(|e| InterpretError::new(format!("User prompt failed: {e}")))?;
if !confirmation {
return Err(InterpretError::new(
"User aborted the operation.".to_string(),
));
}
Ok(())
}
}
#[async_trait]
@@ -74,10 +233,23 @@ impl Interpret<HAClusterTopology> for OKDSetup04WorkersInterpret {
async fn execute(
&self,
_inventory: &Inventory,
_topology: &HAClusterTopology,
inventory: &Inventory,
topology: &HAClusterTopology,
) -> Result<Outcome, InterpretError> {
self.render_and_reboot().await?;
// 1. Ensure we have 2 physical hosts for the worker nodes.
let nodes = self.get_nodes(inventory, topology).await?;
// 2. Create DHCP reservations for the worker nodes.
self.configure_host_binding(inventory, topology, &nodes)
.await?;
// 3. Create iPXE files for each worker node to boot from the worker ignition.
self.configure_ipxe(inventory, topology, &nodes).await?;
// 4. Reboot the nodes to start the OS installation.
self.reboot_targets(&nodes).await?;
Ok(Outcome::success("Workers provisioned".into()))
}
}

View File

@@ -417,6 +417,7 @@ pub struct EthernetSpec {
#[serde(rename_all = "kebab-case")]
pub struct BondSpec {
pub mode: String,
#[serde(alias = "port")]
pub ports: Vec<String>,
#[serde(skip_serializing_if = "Option::is_none")]
pub options: Option<BTreeMap<String, Value>>,

View File

@@ -1,6 +1,6 @@
use async_trait::async_trait;
use harmony_types::id::Id;
use log::{debug, info, warn};
use log::{info, warn};
use serde::Serialize;
use crate::{
@@ -90,7 +90,7 @@ impl HostNetworkConfigurationInterpret {
);
} else {
warn!(
"[Host {current_host}/{total_hosts}] Found a single port for {} interface, skipping",
"[Host {current_host}/{total_hosts}] Found a single port for {} interfaces, skipping",
host.network.len()
);
}
@@ -150,15 +150,6 @@ impl HostNetworkConfigurationInterpret {
];
for config in configs {
let host = self
.score
.hosts
.iter()
.find(|h| h.id == config.host_id)
.unwrap();
println!("[Host] {host}");
if config.switch_ports.is_empty() {
report.push(format!(
"⏭️ Host {}: SKIPPED (No matching switch ports found)",

View File

@@ -1,6 +1,6 @@
use serde::{Deserialize, Serialize};
#[derive(Copy, Clone, Debug, PartialEq, Eq, Hash, Serialize, Deserialize, PartialOrd, Ord)]
#[derive(Copy, Clone, PartialEq, Eq, Hash, Serialize, Deserialize, PartialOrd, Ord)]
pub struct MacAddress(pub [u8; 6]);
impl MacAddress {
@@ -19,6 +19,14 @@ impl From<&MacAddress> for String {
}
}
impl std::fmt::Debug for MacAddress {
fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
f.debug_tuple("MacAddress")
.field(&String::from(self))
.finish()
}
}
impl std::fmt::Display for MacAddress {
fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
f.write_str(&String::from(self))