fix(agent): GPU worker thread, async journalctl, correctness + cache fixes

Lightweight:
- GPU collection moves to a dedicated worker thread that owns the gfxinfo
  handle for the process lifetime. gfxinfo's active_gpu() runs a full NVML
  init/teardown (~20ms, blocking) and we were paying it on the async
  runtime for every collect — measured at ~80% of the agent's entire
  active CPU on a GPU machine. The handle holds Rc<Nvml> (not Send), so a
  thread + mpsc/oneshot channel pair confines it; a zero-total-VRAM reply
  is treated as a dead session (driver reload) and re-probed.
- journalctl now runs via tokio::process instead of blocking one of the
  two runtime workers for the duration of the subprocess.
- TtlCell (state.rs) replaces the four hand-rolled static TTL caches; a
  cached negative result now counts as fresh, so hosts with no matching
  temp sensor or GPU stop rescanning every request. Single lock+clone on
  the GPU cache hit path (was two).

Correctness:
- Process/child CPU times are now microseconds as documented; they were
  milliseconds, rendering 1000x too small next to (correct) thread times.
- Non-Linux per-process CPU%% clamps AFTER dividing by core count; a
  4-cores-busy process on an 8-core box reported 12.5% instead of 50%.
- Journal timestamps are real RFC 3339 UTC plus an additive timestamp_us
  field (sorting is now numeric); the old strings were Debug-formatted
  SystemTime mangled by string replace.
- Partition detection uses /sys/block on Linux: whole-disk filesystems on
  names like nvme0n1 or zram1 are no longer misclassified as partitions.
  One shared parent_disk_name() replaces two inline copies.
- New sampled_at_ms on the metrics payload (additive) records when the
  snapshot was actually collected, so clients can compute exact rates
  across the agent's TTL cache.

Security/robustness:
- key.pem is created 0600 (was umask default 0644, world-readable) and
  pre-1.51 keys are tightened on startup.
- Per-PID detail/journal caches now evict (60s max age, 64 entries max);
  they previously grew without bound under PID-walking clients.
- The two per-PID ws handlers collapse into one generic helper.
- /proc/<pid>/stat parsing unified in one comm-safe module.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
This commit is contained in:
jasonwitty
2026-08-19 14:43:37 -07:00
parent fbc788c799
commit e4b0d9b9f6
13 changed files with 511 additions and 373 deletions
+50 -1
View File
@@ -74,6 +74,55 @@ pub struct AppState {
pub cache_journal_entries: Arc<Mutex<HashMap<u32, CacheEntry<crate::types::JournalResponse>>>>,
}
/// TTL-gated value behind a std Mutex, for `static` caches on hot paths.
/// Replaces the hand-rolled TempCache/GpuCache/refresh-timestamp statics
/// that each reimplemented the same at/value pair.
pub struct TtlCell<T> {
inner: std::sync::Mutex<CacheEntry<T>>,
}
impl<T: Clone> Default for TtlCell<T> {
fn default() -> Self {
Self::new()
}
}
impl<T: Clone> TtlCell<T> {
pub const fn new() -> Self {
Self {
inner: std::sync::Mutex::new(CacheEntry::new()),
}
}
/// The stored value, only while fresh. Poisoned lock reads as a miss.
pub fn get_fresh(&self, ttl: Duration) -> Option<T> {
let g = self.inner.lock().ok()?;
if g.is_fresh(ttl) {
g.value.clone()
} else {
None
}
}
pub fn set(&self, v: T) {
if let Ok(mut g) = self.inner.lock() {
g.set(v);
}
}
/// True exactly once per TTL window: restamps and tells the caller to do
/// the refresh. Atomic check-and-stamp so concurrent callers don't both
/// refresh.
pub fn claim_stale(&self, ttl: Duration) -> bool {
let Ok(mut g) = self.inner.lock() else {
return false;
};
if g.at.is_none_or(|t| t.elapsed() >= ttl) {
g.at = Some(Instant::now());
true
} else {
false
}
}
}
#[derive(Clone, Debug)]
pub struct CacheEntry<T> {
pub at: Option<Instant>,
@@ -87,7 +136,7 @@ impl<T> Default for CacheEntry<T> {
}
impl<T> CacheEntry<T> {
pub fn new() -> Self {
pub const fn new() -> Self {
Self {
at: None,
value: None,