feat(OKDInstallation): Implemented bootstrap of okd worker node, added features to allow both control plane and worker node to use the same bootstrap_okd_node score
All checks were successful
Run Check Script / check (pull_request) Successful in 1m46s

This commit is contained in:
2025-12-10 12:15:07 -05:00
parent 83c1cc82b6
commit 50bd5c5bba
7 changed files with 410 additions and 180 deletions

15
Cargo.lock generated
View File

@@ -6049,6 +6049,21 @@ version = "0.5.1"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "8f50febec83f5ee1df3015341d8bd429f2d1cc62bcba7ea2076759d315084683"
[[package]]
name = "test-score"
version = "0.1.0"
dependencies = [
"base64 0.22.1",
"env_logger",
"harmony",
"harmony_cli",
"harmony_macros",
"harmony_types",
"log",
"tokio",
"url",
]
[[package]]
name = "thiserror"
version = "1.0.69"

View File

@@ -1,4 +1,6 @@
mod repository;
use std::fmt;
pub use repository::*;
#[derive(Debug, new, Clone)]
@@ -71,3 +73,14 @@ pub enum HostRole {
Worker,
Storage,
}
impl fmt::Display for HostRole {
fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
match self {
HostRole::Bootstrap => write!(f, "Bootstrap"),
HostRole::ControlPlane => write!(f, "ControlPlane"),
HostRole::Worker => write!(f, "Worker"),
HostRole::Storage => write!(f, "Storage"),
}
}
}

View File

@@ -5,8 +5,10 @@ use crate::{
interpret::{Interpret, InterpretError, InterpretName, InterpretStatus, Outcome},
inventory::{HostRole, Inventory},
modules::{
dhcp::DhcpHostBindingScore, http::IPxeMacBootFileScore,
inventory::DiscoverHostForRoleScore, okd::templates::BootstrapIpxeTpl,
dhcp::DhcpHostBindingScore,
http::IPxeMacBootFileScore,
inventory::DiscoverHostForRoleScore,
okd::{bootstrap_okd_node::OKDNodeInterpret, templates::BootstrapIpxeTpl},
},
score::Score,
topology::{HAClusterTopology, HostBinding},
@@ -50,159 +52,6 @@ impl OKDSetup03ControlPlaneInterpret {
status: InterpretStatus::QUEUED,
}
}
/// 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 = 3;
let repo = InventoryRepositoryFactory::build().await?;
let mut control_plane_hosts = repo.get_host_for_role(&HostRole::ControlPlane).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::ControlPlane,
}
.interpret(inventory, topology)
.await?;
control_plane_hosts = repo.get_host_for_role(&HostRole::ControlPlane).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!("[ControlPlane] Configuring host bindings for control plane 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!("[ControlPlane] 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: "master.ign", // Control plane nodes use the master ignition file
}
.to_string();
debug!("[ControlPlane] 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!(
"[ControlPlane] 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!("[ControlPlane] Requesting reboot for control plane nodes: {node_ids:?}",);
let confirmation = inquire::Confirm::new(
&format!("Please reboot the {} control plane 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]
@@ -228,19 +77,10 @@ impl Interpret<HAClusterTopology> for OKDSetup03ControlPlaneInterpret {
inventory: &Inventory,
topology: &HAClusterTopology,
) -> Result<Outcome, InterpretError> {
// 1. Ensure we have 3 physical hosts for the control plane.
let nodes = self.get_nodes(inventory, topology).await?;
// 2. Create DHCP reservations for the control plane nodes.
self.configure_host_binding(inventory, topology, &nodes)
OKDNodeInterpret::new(HostRole::ControlPlane)
.execute(inventory, topology)
.await?;
// 3. Create iPXE files for each control plane node to boot from the master ignition.
self.configure_ipxe(inventory, topology, &nodes).await?;
// 4. Reboot the nodes to start the OS installation.
self.reboot_targets(&nodes).await?;
// TODO: Implement a step to wait for the control plane nodes to join the cluster
// and for the cluster operators to become available. This would be similar to
// the `wait-for bootstrap-complete` command.

View File

@@ -1,13 +1,13 @@
use async_trait::async_trait;
use derive_new::new;
use harmony_types::id::Id;
use log::info;
use serde::Serialize;
use crate::{
data::Version,
interpret::{Interpret, InterpretError, InterpretName, InterpretStatus, Outcome},
inventory::Inventory,
inventory::{HostRole, Inventory},
modules::okd::bootstrap_okd_node::OKDNodeInterpret,
score::Score,
topology::HAClusterTopology,
};
@@ -23,7 +23,7 @@ pub struct OKDSetup04WorkersScore {}
impl Score<HAClusterTopology> for OKDSetup04WorkersScore {
fn create_interpret(&self) -> Box<dyn Interpret<HAClusterTopology>> {
Box::new(OKDSetup04WorkersInterpret::new(self.clone()))
Box::new(OKDSetup04WorkersInterpret::new())
}
fn name(&self) -> String {
@@ -33,25 +33,18 @@ impl Score<HAClusterTopology> for OKDSetup04WorkersScore {
#[derive(Debug, Clone)]
pub struct OKDSetup04WorkersInterpret {
score: OKDSetup04WorkersScore,
version: Version,
status: InterpretStatus,
}
impl OKDSetup04WorkersInterpret {
pub fn new(score: OKDSetup04WorkersScore) -> Self {
pub fn new() -> Self {
let version = Version::from("1.0.0").unwrap();
Self {
version,
score,
status: InterpretStatus::QUEUED,
}
}
async fn render_and_reboot(&self) -> Result<(), InterpretError> {
info!("[Workers] Rendering per-MAC PXE for workers and rebooting");
Ok(())
}
}
#[async_trait]
@@ -74,10 +67,12 @@ 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?;
OKDNodeInterpret::new(HostRole::Worker)
.execute(inventory, topology)
.await?;
Ok(Outcome::success("Workers provisioned".into()))
}
}

View File

@@ -0,0 +1,296 @@
use async_trait::async_trait;
use derive_new::new;
use harmony_types::id::Id;
use log::{debug, info};
use serde::Serialize;
use crate::{
data::Version,
hardware::PhysicalHost,
infra::inventory::InventoryRepositoryFactory,
interpret::{Interpret, InterpretError, InterpretName, InterpretStatus, Outcome},
inventory::{HostRole, Inventory},
modules::{
dhcp::DhcpHostBindingScore,
http::IPxeMacBootFileScore,
inventory::DiscoverHostForRoleScore,
okd::{
okd_node::{
BootstrapRole, ControlPlaneRole, OKDRoleProperties, StorageRole, WorkerRole,
},
templates::BootstrapIpxeTpl,
},
},
score::Score,
topology::{HAClusterTopology, HostBinding, LogicalHost},
};
#[derive(Debug, Clone, Serialize, new)]
pub struct OKDNodeScore {
host_role: HostRole,
}
impl Score<HAClusterTopology> for OKDNodeScore {
fn name(&self) -> String {
"OKDNodeScore".to_string()
}
fn create_interpret(&self) -> Box<dyn Interpret<HAClusterTopology>> {
Box::new(OKDNodeInterpret::new(self.host_role.clone()))
}
}
#[derive(Debug, Clone)]
pub struct OKDNodeInterpret {
host_role: HostRole,
}
impl OKDNodeInterpret {
pub fn new(host_role: HostRole) -> Self {
Self { host_role }
}
fn okd_role_properties(&self, role: &HostRole) -> &'static dyn OKDRoleProperties {
match role {
HostRole::Bootstrap => &BootstrapRole,
HostRole::ControlPlane => &ControlPlaneRole,
HostRole::Worker => &WorkerRole,
HostRole::Storage => &StorageRole,
}
}
async fn get_nodes(
&self,
inventory: &Inventory,
topology: &HAClusterTopology,
) -> Result<Vec<PhysicalHost>, InterpretError> {
let repo = InventoryRepositoryFactory::build().await?;
let mut hosts = repo.get_host_for_role(&self.host_role).await?;
let okd_host_properties = self.okd_role_properties(&self.host_role);
let required_hosts: usize = okd_host_properties.required_hosts();
while hosts.len() < required_hosts {
info!(
"Discovery of {} {} hosts in progress, current number {}",
required_hosts,
self.host_role,
hosts.len()
);
// This score triggers the discovery agent for a specific role.
DiscoverHostForRoleScore {
role: self.host_role.clone(),
}
.interpret(inventory, topology)
.await?;
hosts = repo.get_host_for_role(&self.host_role).await?;
}
if hosts.len() < required_hosts {
Err(InterpretError::new(format!(
"OKD Requires at least {} {} hosts, but only found {}. Cannot proceed.",
required_hosts,
self.host_role,
hosts.len()
)))
} else {
// Take exactly the number of required hosts to ensure consistency.
Ok(hosts.into_iter().take(required_hosts).collect())
}
}
/// Configures DHCP host bindings for all nodes.
async fn configure_host_binding(
&self,
inventory: &Inventory,
topology: &HAClusterTopology,
nodes: &Vec<PhysicalHost>,
) -> Result<(), InterpretError> {
info!(
"[{}] Configuring host bindings for {} plane nodes.",
self.host_role, self.host_role,
);
let host_properties = self.okd_role_properties(&self.host_role);
self.validate_host_node_match(nodes, host_properties.logical_hosts(topology))?;
let bindings: Vec<HostBinding> =
self.host_bindings(nodes, host_properties.logical_hosts(topology));
DhcpHostBindingScore {
host_binding: bindings,
domain: Some(topology.domain_name.clone()),
}
.interpret(inventory, topology)
.await?;
Ok(())
}
// Ensure the topology definition matches the number of physical nodes found.
fn validate_host_node_match(
&self,
nodes: &Vec<PhysicalHost>,
hosts: &Vec<LogicalHost>,
) -> Result<(), InterpretError> {
if hosts.len() != nodes.len() {
return Err(InterpretError::new(format!(
"Mismatch between logical hosts defined in topology ({}) and physical nodes found ({}).",
hosts.len(),
nodes.len()
)));
}
Ok(())
}
// Create a binding for each physical host to its corresponding logical host.
fn host_bindings(
&self,
nodes: &Vec<PhysicalHost>,
hosts: &Vec<LogicalHost>,
) -> Vec<HostBinding> {
hosts
.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()
}
/// Renders and deploys a per-MAC iPXE boot file for each node.
async fn configure_ipxe(
&self,
inventory: &Inventory,
topology: &HAClusterTopology,
nodes: &Vec<PhysicalHost>,
) -> Result<(), InterpretError> {
info!(
"[{}] Rendering per-MAC iPXE configurations.",
self.host_role
);
let okd_role_properties = self.okd_role_properties(&self.host_role);
// 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",
//TODO must be refactored to not only use /dev/sda
installation_device: "/dev/sda", // This might need to be configurable per-host in the future
ignition_file_name: okd_role_properties.ignition_file(),
}
.to_string();
debug!("[{}] iPXE content template:\n{content}", self.host_role);
// 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!(
"[{}] Applying iPXE config for node ID '{}' with MACs: {:?}",
self.host_role, 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!(
"[{}] Requesting reboot for control plane nodes: {node_ids:?}",
self.host_role
);
let confirmation = inquire::Confirm::new(
&format!("Please reboot the {} {} nodes ({}) to apply their PXE configuration. Press enter when ready.", nodes.len(), self.host_role, 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]
impl Interpret<HAClusterTopology> for OKDNodeInterpret {
async fn execute(
&self,
inventory: &Inventory,
topology: &HAClusterTopology,
) -> Result<Outcome, InterpretError> {
// 1. Ensure we have the specfied number of physical hosts.
let nodes = self.get_nodes(inventory, topology).await?;
// 2. Create DHCP reservations for the nodes.
self.configure_host_binding(inventory, topology, &nodes)
.await?;
// 3. Create iPXE files for each node to boot from the ignition.
self.configure_ipxe(inventory, topology, &nodes).await?;
// 4. Reboot the nodes to start the OS installation.
self.reboot_targets(&nodes).await?;
// TODO: Implement a step to wait for the control plane nodes to join the cluster
// and for the cluster operators to become available. This would be similar to
// the `wait-for bootstrap-complete` command.
info!(
"[{}] Provisioning initiated. Monitor the cluster convergence manually.",
self.host_role
);
Ok(Outcome::success(format!(
"{} provisioning has been successfully initiated.",
self.host_role
)))
}
fn get_name(&self) -> InterpretName {
InterpretName::Custom("OKDNodeSetup".into())
}
fn get_version(&self) -> Version {
todo!()
}
fn get_status(&self) -> InterpretStatus {
todo!()
}
fn get_children(&self) -> Vec<Id> {
todo!()
}
}

View File

@@ -6,12 +6,14 @@ mod bootstrap_05_sanity_check;
mod bootstrap_06_installation_report;
pub mod bootstrap_dhcp;
pub mod bootstrap_load_balancer;
pub mod bootstrap_okd_node;
mod bootstrap_persist_network_bond;
pub mod dhcp;
pub mod dns;
pub mod installation;
pub mod ipxe;
pub mod load_balancer;
pub mod okd_node;
pub mod templates;
pub mod upgrade;
pub use bootstrap_01_prepare::*;

View File

@@ -0,0 +1,69 @@
use crate::topology::{HAClusterTopology, LogicalHost};
pub trait OKDRoleProperties {
fn ignition_file(&self) -> &'static str;
fn required_hosts(&self) -> usize;
fn logical_hosts<'a>(&self, t: &'a HAClusterTopology) -> &'a Vec<LogicalHost>;
}
pub struct BootstrapRole;
pub struct ControlPlaneRole;
pub struct WorkerRole;
pub struct StorageRole;
impl OKDRoleProperties for BootstrapRole {
fn ignition_file(&self) -> &'static str {
"bootstrap.ign"
}
fn required_hosts(&self) -> usize {
1
}
fn logical_hosts<'a>(&self, t: &'a HAClusterTopology) -> &'a Vec<LogicalHost> {
todo!()
}
}
impl OKDRoleProperties for ControlPlaneRole {
fn ignition_file(&self) -> &'static str {
"master.ign"
}
fn required_hosts(&self) -> usize {
3
}
fn logical_hosts<'a>(&self, t: &'a HAClusterTopology) -> &'a Vec<LogicalHost> {
&t.control_plane
}
}
impl OKDRoleProperties for WorkerRole {
fn ignition_file(&self) -> &'static str {
"worker.ign"
}
fn required_hosts(&self) -> usize {
2
}
fn logical_hosts<'a>(&self, t: &'a HAClusterTopology) -> &'a Vec<LogicalHost> {
&t.workers
}
}
//TODO unsure if this is to be implemented here or not
impl OKDRoleProperties for StorageRole {
fn ignition_file(&self) -> &'static str {
todo!()
}
fn required_hosts(&self) -> usize {
todo!()
}
fn logical_hosts<'a>(&self, t: &'a HAClusterTopology) -> &'a Vec<LogicalHost> {
todo!()
}
}