feat(jobs): add admin call to purge jubs result

This commit is contained in:
Edouard Vanbelle
2026-07-30 00:21:29 +02:00
parent 507bc2e98d
commit 9fe47a53ea
6 changed files with 316 additions and 5 deletions
@@ -431,6 +431,30 @@ impl JobStoreProvider for PgJobStoreProvider {
.collect())
}
async fn purge_terminal_runs(&self, retention_days: i32) -> Result<u64, DomainError> {
// Defensive floor — zero would eat just-completed runs;
// negative would eat the whole terminal history.
let days = retention_days.max(1);
// `ON DELETE CASCADE` on jobs.run_findings.run_id
// (migration 20260930000001) drops findings with their
// parent run. Non-terminal statuses (Running / Paused /
// CancelRequested) explicitly excluded to protect
// in-flight work.
let result = sqlx::query(
r#"
DELETE FROM jobs.recoverable_runs
WHERE status IN ('Completed', 'Failed')
AND completed_at IS NOT NULL
AND completed_at < NOW() - ($1 || ' days')::interval
"#,
)
.bind(days.to_string())
.execute(self.pool.as_ref())
.await
.map_err(|e| map_sqlx_err("purge_terminal_runs", e))?;
Ok(result.rows_affected())
}
async fn request_cancel(&self, job_name: &str) -> Result<Option<Uuid>, DomainError> {
// Only Running → CancelRequested flips. `Paused` can be
// cancelled by not resuming — no need for a state change.
@@ -371,6 +371,29 @@ pub trait JobStoreProvider: Send + Sync {
&self,
run_id: Uuid,
) -> Result<Vec<(String, u64)>, DomainError>;
/// Operator-triggered retention cleanup. DELETEs every
/// TERMINAL run (`Completed`, `Failed`) whose `completed_at`
/// is older than `retention_days` days ago. Findings drop
/// alongside via the `ON DELETE CASCADE` FK on
/// `jobs.run_findings.run_id`.
///
/// Non-terminal rows (`Running`, `Paused`, `CancelRequested`)
/// are ALWAYS preserved regardless of age — an in-flight or
/// paused run must not be reaped by retention.
///
/// `retention_days` is treated as `max(1, retention_days)` at
/// the impl layer to defend against a zero/negative value
/// eating just-completed runs.
///
/// Returns the number of run rows deleted (which equals
/// the number of finding rows deleted *transitively* via
/// CASCADE; callers wanting the finding count separately
/// should query it BEFORE calling this).
///
/// Powers `POST /api/admin/jobs/runs/purge`. Not periodic — the
/// operator decides when to reclaim space.
async fn purge_terminal_runs(&self, retention_days: i32) -> Result<u64, DomainError>;
}
/// How a `RunProgress` fraction was derived. Lets the UI communicate
@@ -1094,6 +1117,25 @@ mod tests {
Ok(counts.into_iter().collect())
}
async fn purge_terminal_runs(&self, retention_days: i32) -> Result<u64, DomainError> {
// Test-double: no `completed_at` to compare against, so
// just drop every terminal-state store when
// `retention_days` > 0. Sufficient for the trait
// contract check; PG impl exercises the real
// `completed_at < NOW() - days` filter.
let days = retention_days.max(1);
if days == 0 {
return Ok(0);
}
let mut stores = self.stores.lock().unwrap();
let before = stores.len();
stores.retain(|s| {
let state = s.state.lock().unwrap();
!matches!(state.status, RunStatus::Completed | RunStatus::Failed)
});
Ok((before - stores.len()) as u64)
}
async fn request_cancel(&self, _job_name: &str) -> Result<Option<Uuid>, DomainError> {
let stores = self.stores.lock().unwrap();
if let Some(s) = stores.last() {