git-echo/linux_patch_manager
Private
Public Access
1
0
Fork 0
Code Releases 8 Wiki Activity

feat: M6 maintenance windows + M7 WebSocket relay (real-time job status)

Browse Source 
M6 - Maintenance Windows:
- routes/maintenance_windows.rs: full CRUD API
- migrations/004_maintenance_windows.sql
- frontend/MaintenanceWindowsPage.tsx
- HostDetailPage.tsx: maintenance window config panel

M7 - WebSocket Relay:
- pm-web: POST /api/v1/ws/ticket (JWT-auth, single-use, 60s TTL)
- pm-web: WS /api/v1/ws/jobs?ticket=... (PgListener -> browser push)
- pm-web: DashMap<String,WsTicket> in AppState, 30s cleanup task
- pm-worker: ws_relay.rs subscribes to agent WS, updates patch_job_hosts,
  fires pg_notify(job_update) for real-time fan-out
- frontend: useJobWebSocket hook with auto-reconnect + exponential backoff
- frontend: JobsPage live updates with WS status indicator
- types: JobWsEvent interface
- api/client: wsApi.createTicket()

All tasks marked complete in tasks/todo.md
cargo build: zero errors, zero warnings
This commit is contained in:
Echo
2026-04-23 17:42:51 +00:00
parent 6f9c6dc881
commit a5d52ffab0
21 changed files with 2833 additions and 36 deletions
Download Patch File Download Diff File Expand all files Collapse all files

35
crates/pm-worker/src/job_executor.rs
View File

@ -204,6 +204,39 @@ async fn scan_queued_jobs(pool: PgPool, config: Arc<AppConfig>) {
WHERE pjh.status = 'queued'
AND (pjh.retry_next_at IS NULL OR pjh.retry_next_at <= NOW())
AND j.status != 'cancelled'
AND (
-- Immediate jobs always dispatch
j.immediate = TRUE
OR
-- Non-immediate jobs only dispatch when the host has an open window
EXISTS (
SELECT 1 FROM maintenance_windows mw
WHERE mw.host_id = pjh.host_id
AND mw.enabled = TRUE
AND (
(mw.recurrence = 'once'
AND mw.start_at <= NOW()
AND NOW() < mw.start_at + (mw.duration_minutes * INTERVAL '1 minute'))
OR
(mw.recurrence = 'daily'
AND (NOW() AT TIME ZONE 'UTC')::time >= (mw.start_at AT TIME ZONE 'UTC')::time
AND (NOW() AT TIME ZONE 'UTC')::time < ((mw.start_at AT TIME ZONE 'UTC')::time
+ (mw.duration_minutes * INTERVAL '1 minute')))
OR
(mw.recurrence = 'weekly'
AND EXTRACT(DOW FROM NOW() AT TIME ZONE 'UTC') = mw.recurrence_day
AND (NOW() AT TIME ZONE 'UTC')::time >= (mw.start_at AT TIME ZONE 'UTC')::time
AND (NOW() AT TIME ZONE 'UTC')::time < ((mw.start_at AT TIME ZONE 'UTC')::time
+ (mw.duration_minutes * INTERVAL '1 minute')))
OR
(mw.recurrence = 'monthly'
AND EXTRACT(DAY FROM NOW() AT TIME ZONE 'UTC') = mw.recurrence_day
AND (NOW() AT TIME ZONE 'UTC')::time >= (mw.start_at AT TIME ZONE 'UTC')::time
AND (NOW() AT TIME ZONE 'UTC')::time < ((mw.start_at AT TIME ZONE 'UTC')::time
+ (mw.duration_minutes * INTERVAL '1 minute')))
)
)
)
"#,
)
.fetch_all(&pool)
@ -230,7 +263,7 @@ async fn scan_queued_jobs(pool: PgPool, config: Arc<AppConfig>) {
/// Fetch all queued host entries for `job_id` and dispatch them concurrently,
/// bounded by `config.worker.max_concurrent_agent_calls`.
async fn process_job(pool: PgPool, config: Arc<AppConfig>, job_id: Uuid) {
pub async fn process_job(pool: PgPool, config: Arc<AppConfig>, job_id: Uuid) {
tracing::info!(%job_id, "process_job: dispatching queued hosts");
// Mark the parent job as running (idempotent guard).