fix(analysis): emit Cancelled event + assert lock-release after cancel (0.87.13)

Two 0.87.4 follow-ups:

1. Cron tick: existing shutdown-on-cancel test now also asserts the
   advisory lock is re-acquirable after cancel-during-tick. Without
   this, a refactor moving `pg_advisory_unlock` under the cancel arm
   would ship green and break multi-replica safety.

2. Analysis worker: cancel-mid-dispatch left the dashboard banner stuck
   on the cancelled page until a later page overwrote it. Add a
   `Cancelled` AnalysisEvent variant, publish on the cancel-release
   branch, wire the analysis and admin overview dashboards to clear
   on it. Per-page chip rolls back from `analyzing` to `queued`.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
This commit is contained in:
MechaCat02
2026-06-23 08:01:53 +02:00
parent 95b98eebf1
commit 651fb27522
9 changed files with 88 additions and 5 deletions

2
backend/Cargo.lock generated
View File

@@ -1517,7 +1517,7 @@ checksum = "c41e0c4fef86961ac6d6f8a82609f55f31b05e4fce149ac5710e439df7619ba4"
[[package]]
name = "mangalord"
version = "0.87.12"
version = "0.87.13"
dependencies = [
"anyhow",
"argon2",

View File

@@ -1,6 +1,6 @@
[package]
name = "mangalord"
version = "0.87.12"
version = "0.87.13"
edition = "2021"
default-run = "mangalord"

View File

@@ -276,7 +276,23 @@ impl WorkerContext {
);
let _ = jobs::release(&self.pool, lease.id).await;
// Don't publish Failed (the page didn't actually fail) and
// don't write a duration row — neither outcome is true.
// don't write a duration row — neither outcome is true. DO
// publish Cancelled so the dashboard's "now analyzing" banner
// can clear the page we previously sent Started for. Without
// this the banner stuck on the cancelled page until a later
// page kicked off and overwrote it.
if let Some((manga_id, manga_title, chapter_id, chapter_number, page_number)) =
breadcrumb
{
self.events.publish(AnalysisEvent::Cancelled {
page_id,
manga_id,
manga_title,
chapter_id,
chapter_number,
page_number,
});
}
return;
};

View File

@@ -54,6 +54,20 @@ pub enum AnalysisEvent {
chapter_number: i32,
page_number: i32,
},
/// The worker cancelled an in-flight dispatch (daemon shutdown or
/// settings reload toggling analysis off). The lease was released,
/// not failed — but the dashboard banner already saw `Started` and
/// only clears on `Completed`/`Failed`/`Cancelled`. Without this
/// variant the banner stuck on the cancelled page forever (until a
/// later page kicked off and overwrote it).
Cancelled {
page_id: Uuid,
manga_id: Uuid,
manga_title: String,
chapter_id: Uuid,
chapter_number: i32,
page_number: i32,
},
}
/// Broadcaster shared on `AppState`. Cheap to clone via `Arc`. Publishing

View File

@@ -894,4 +894,27 @@ async fn cron_shutdown_does_not_block_on_in_flight_metadata_pass(pool: PgPool) {
started.elapsed() < Duration::from_secs(5),
"shutdown took too long; cancel didn't propagate into the cron tick body"
);
// The advisory lock must be released so another replica's cron can
// pick up the slot. Without this assertion, a refactor that moved
// the `pg_advisory_unlock` under the cancel arm (instead of running
// unconditionally below the select!) would ship green — only to
// wedge multi-replica deployments where the second replica's cron
// would block on `pg_try_advisory_lock` forever. `pg_try_advisory_lock`
// returns true iff the lock is free, so we use it as the probe.
let acquired: bool = sqlx::query_scalar("SELECT pg_try_advisory_lock($1)")
.bind(CRON_LOCK_KEY)
.fetch_one(&pool)
.await
.unwrap();
assert!(
acquired,
"advisory lock must be released after cancel-during-tick \
(otherwise multi-replica safety breaks)"
);
// Be a good citizen for whatever sqlx test runs next.
let _ = sqlx::query("SELECT pg_advisory_unlock($1)")
.bind(CRON_LOCK_KEY)
.execute(&pool)
.await;
}

View File

@@ -1,6 +1,6 @@
{
"name": "mangalord-frontend",
"version": "0.87.12",
"version": "0.87.13",
"private": true,
"type": "module",
"scripts": {

View File

@@ -857,7 +857,12 @@ export type AnalysisEvent =
chapter_id: string | null;
}
| {
kind: 'started' | 'completed' | 'failed';
// `cancelled` fires when the worker drops an in-flight dispatch
// on daemon shutdown / settings-toggle-off. The lease was
// released, not failed — the page returns to pending. The
// dashboard banner clears on this so it doesn't stick on the
// cancelled page until another `started` overwrites it.
kind: 'started' | 'completed' | 'failed' | 'cancelled';
page_id: string;
manga_id: string;
manga_title: string;

View File

@@ -121,6 +121,16 @@
function applyAnalysisEvent(ev: AnalysisEvent) {
if (ev.kind === 'enqueued') return;
// Cancelled (daemon shutdown / settings toggle-off mid-dispatch)
// is a non-terminal release — the lease went back to pending.
// Clear the banner instead of frame-flipping to a "done"/"failed"
// misrepresentation.
if (ev.kind === 'cancelled') {
if (nowAnalyzing && nowAnalyzing.page === ev.page_number) {
nowAnalyzing = null;
}
return;
}
const phase =
ev.kind === 'started' ? 'analyzing' : ev.kind === 'completed' ? 'done' : 'failed';
nowAnalyzing = {

View File

@@ -247,6 +247,21 @@
scheduleClear(ev.page_id);
}
pushTicker(`✗ Failed page ${ev.page_number}`);
} else if (ev.kind === 'cancelled') {
// Daemon shutdown / analysis toggled off mid-dispatch. The
// lease was released, not failed — the page is still pending
// server-side. We just need to clear the "Analyzing pN" banner
// and roll the per-page live status back to queued so the
// chip doesn't stay spinner-frozen forever.
const cur = liveStatus[ev.page_id];
if (cur === 'analyzing') {
liveStatus = { ...liveStatus, [ev.page_id]: 'queued' };
}
if (nowAnalyzing?.page_id === ev.page_id) {
nowAnalyzing = null;
if (clearTimer) clearTimeout(clearTimer);
}
pushTicker(`↺ Cancelled page ${ev.page_number}`);
}
}