feat(host-agent): Rust rewrite Phase 0 — multi-instance foundation, v2 wire protocol, real telemetry
All checks were successful
Test Asgard Runner / test (push) Successful in 3s

New corrosion-host-agent/ crate (Go companion-agent stays as behavior
reference until parity). Wire protocol v2 per COA-B: instance-scoped
subjects corrosion.{license}.{instance}.* + host-level .host.* — spec
in PROTOCOL.md, designed for the license->host->instance fleet model.

- Multi-instance TOML config in the foundation, not retrofitted
- NATS layer on the Vigilance production profile (infinite reconnect,
  capped backoff, 30s ping, 8192-msg offline buffer)
- Heartbeat with real sysinfo telemetry — Go agent shipped hardcoded
  disk/cpu placeholders; this is the panel's first true Resources data
- Connectivity prober (outbound TCP, periodic + on-demand)
- Host cmd channel (ping/probe/sysinfo), going-offline beacon,
  CancellationToken shutdown
- Live-fire verified against production NATS; artifacts: 3.7MB static
  linux-musl, 3.8MB windows .exe (static CRT)

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
This commit is contained in:
Vantz Stockwell
2026-06-11 10:02:46 -04:00
parent 1abe57ca40
commit cea3d66cdd
18 changed files with 3292 additions and 0 deletions

View File

@@ -0,0 +1,186 @@
//! Agent configuration: TOML file + environment overrides.
//!
//! Multi-instance is foundational, not bolted on: one agent supervises N game
//! instances on the host, each declared as an `[[instance]]` block. Connection
//! secrets may come from env so the config file can be world-readable-ish
//! while the token is not.
use anyhow::{bail, Context, Result};
use serde::Deserialize;
use std::collections::HashSet;
use std::path::{Path, PathBuf};
/// Instance ids share the NATS subject namespace with host-level segments.
const RESERVED_INSTANCE_IDS: &[&str] = &["host", "cmd", "files", "update", "agent"];
pub const SUPPORTED_GAMES: &[&str] = &["rust", "conan", "soulmask", "dune"];
#[derive(Debug, Clone, Deserialize)]
#[serde(deny_unknown_fields)]
pub struct ConfigFile {
pub agent: AgentSection,
#[serde(default, rename = "instance")]
pub instances: Vec<InstanceConfig>,
#[serde(default)]
pub prober: ProberSection,
}
#[derive(Debug, Clone, Deserialize)]
#[serde(deny_unknown_fields)]
pub struct AgentSection {
pub license_id: Option<String>,
pub nats_url: Option<String>,
pub nats_token: Option<String>,
#[serde(default = "default_heartbeat_seconds")]
pub heartbeat_seconds: u64,
#[serde(default = "default_log_level")]
pub log_level: String,
}
#[derive(Debug, Clone, Deserialize)]
#[serde(deny_unknown_fields)]
pub struct InstanceConfig {
/// Short slug, unique per license: becomes a NATS subject segment.
pub id: String,
/// One of SUPPORTED_GAMES.
pub game: String,
/// Install root for this instance on the host.
pub root: PathBuf,
/// Optional human label shown in the panel.
#[serde(default)]
pub label: Option<String>,
}
#[derive(Debug, Clone, Default, Deserialize)]
#[serde(deny_unknown_fields)]
pub struct ProberSection {
#[serde(default = "default_probe_interval")]
pub interval_seconds: u64,
/// Extra TCP targets beyond the built-in defaults.
#[serde(default, rename = "target")]
pub targets: Vec<ProbeTargetConfig>,
}
#[derive(Debug, Clone, Deserialize)]
#[serde(deny_unknown_fields)]
pub struct ProbeTargetConfig {
pub name: String,
pub host: String,
pub port: u16,
}
fn default_heartbeat_seconds() -> u64 {
60
}
fn default_probe_interval() -> u64 {
300
}
fn default_log_level() -> String {
"info".to_string()
}
/// Fully-resolved settings after merging file + env. Everything required is
/// present and validated.
#[derive(Debug, Clone)]
pub struct Settings {
pub license_id: String,
pub nats_url: String,
pub nats_token: Option<String>,
pub heartbeat_seconds: u64,
pub log_level: String,
pub instances: Vec<InstanceConfig>,
pub probe_interval_seconds: u64,
pub probe_targets: Vec<ProbeTargetConfig>,
}
pub fn default_config_path() -> PathBuf {
#[cfg(windows)]
{
PathBuf::from(r"C:\ProgramData\Corrosion\agent.toml")
}
#[cfg(not(windows))]
{
PathBuf::from("/etc/corrosion/agent.toml")
}
}
pub fn load(path: &Path) -> Result<Settings> {
let raw = std::fs::read_to_string(path)
.with_context(|| format!("reading config file {}", path.display()))?;
let file: ConfigFile = toml::from_str(&raw)
.with_context(|| format!("parsing config file {}", path.display()))?;
resolve(file)
}
/// Merge env overrides (env wins) and validate.
fn resolve(file: ConfigFile) -> Result<Settings> {
let license_id = std::env::var("CORROSION_LICENSE_ID")
.ok()
.filter(|v| !v.is_empty())
.or(file.agent.license_id)
.context("license_id missing: set [agent].license_id or CORROSION_LICENSE_ID")?;
let nats_url = std::env::var("CORROSION_NATS_URL")
.ok()
.filter(|v| !v.is_empty())
.or(file.agent.nats_url)
.context("nats_url missing: set [agent].nats_url or CORROSION_NATS_URL")?;
let nats_token = std::env::var("CORROSION_NATS_TOKEN")
.ok()
.filter(|v| !v.is_empty())
.or(file.agent.nats_token);
validate_subject_segment("license_id", &license_id)?;
let mut seen: HashSet<&str> = HashSet::new();
for inst in &file.instances {
validate_subject_segment("instance id", &inst.id)?;
if RESERVED_INSTANCE_IDS.contains(&inst.id.as_str()) {
bail!("instance id '{}' is reserved", inst.id);
}
if !seen.insert(inst.id.as_str()) {
bail!("duplicate instance id '{}'", inst.id);
}
if !SUPPORTED_GAMES.contains(&inst.game.as_str()) {
bail!(
"instance '{}': unsupported game '{}' (supported: {})",
inst.id,
inst.game,
SUPPORTED_GAMES.join(", ")
);
}
}
if file.agent.heartbeat_seconds < 10 {
bail!("[agent].heartbeat_seconds must be >= 10");
}
Ok(Settings {
license_id,
nats_url,
nats_token,
heartbeat_seconds: file.agent.heartbeat_seconds,
log_level: file.agent.log_level,
instances: file.instances,
probe_interval_seconds: file.prober.interval_seconds.max(30),
probe_targets: file.prober.targets,
})
}
/// NATS subject segments must not contain '.', '*', '>', whitespace, etc.
/// Keep it strict: lowercase alphanumerics plus '-' and '_', max 64 chars.
fn validate_subject_segment(what: &str, value: &str) -> Result<()> {
if value.is_empty() || value.len() > 64 {
bail!("{what} '{value}' must be 1-64 characters");
}
if !value
.chars()
.all(|c| c.is_ascii_lowercase() || c.is_ascii_digit() || c == '-' || c == '_')
{
bail!("{what} '{value}' may only contain lowercase letters, digits, '-' and '_'");
}
Ok(())
}