feat(jobs): add admin call to purge jubs result
This commit is contained in:
Edouard Vanbelle
@@ -114,6 +114,43 @@ export function listRuns(name: string, limit = 20): Promise<RunSummary[]> {
|
||||
});
|
||||
}
|
||||
|
||||
/** Envelope from `POST /api/admin/jobs/runs/purge`. `purged` is
|
||||
* the count of terminal-run rows deleted (findings cascade with
|
||||
* their parent run via the FK, no separate counter). */
|
||||
export interface PurgeResponse {
|
||||
purged: number;
|
||||
retention_days: number;
|
||||
}
|
||||
|
||||
/**
|
||||
* `POST /api/admin/jobs/runs/purge?days=N` — operator-triggered
|
||||
* retention cleanup. Deletes terminal runs (`Completed`, `Failed`)
|
||||
* with `completed_at` older than `days` days ago; associated
|
||||
* `jobs.run_findings` rows drop with them via CASCADE. Non-terminal
|
||||
* runs (`Running`, `Paused`, `CancelRequested`) are ALWAYS
|
||||
* preserved regardless of age.
|
||||
*
|
||||
* Backend enforces a minimum of 1 day defensively.
|
||||
*/
|
||||
export async function purgeJobRuns(days = 30): Promise<PurgeResponse> {
|
||||
const res = await apiFetch(`/api/admin/jobs/runs/purge?days=${days}`, {
|
||||
method: 'POST',
|
||||
credentials: 'same-origin',
|
||||
headers: { ...JSON_HEADERS, ...getCsrfHeaders() }
|
||||
});
|
||||
if (!res.ok) {
|
||||
let msg = `purge failed: ${res.status}`;
|
||||
try {
|
||||
const body = (await res.json()) as { error?: string; message?: string };
|
||||
msg = body.error ?? body.message ?? msg;
|
||||
} catch {
|
||||
/* no JSON body */
|
||||
}
|
||||
throw new Error(msg);
|
||||
}
|
||||
return (await res.json()) as PurgeResponse;
|
||||
}
|
||||
|
||||
/**
|
||||
* `GET /api/admin/jobs/{name}/runs/{id}/findings?limit=N&offset=M` —
|
||||
* paginated findings for a specific run. Empty list = clean run,
|
||||
|
||||
@@ -20,6 +20,7 @@
|
||||
<script lang="ts">
|
||||
import { SvelteSet } from 'svelte/reactivity';
|
||||
import Icon from '$lib/icons/Icon.svelte';
|
||||
import Modal from '$lib/components/Modal.svelte';
|
||||
import { t } from '$lib/i18n/index.svelte';
|
||||
import { errorMessage } from '$lib/utils/errors';
|
||||
import { ui } from '$lib/stores/ui.svelte';
|
||||
@@ -28,7 +29,8 @@
|
||||
listRuns,
|
||||
listFindings,
|
||||
triggerJob,
|
||||
cancelJob
|
||||
cancelJob,
|
||||
purgeJobRuns
|
||||
} from '$lib/api/endpoints/adminJobs';
|
||||
import type { Finding, JobSummary, RunSummary, RunStatus } from '$lib/api/types';
|
||||
|
||||
@@ -63,6 +65,40 @@
|
||||
else busyKeys.delete(key);
|
||||
}
|
||||
|
||||
// Purge-modal state. Null = closed; otherwise carries the
|
||||
// draft retention days the operator's picking. Kept separate
|
||||
// from the top-bar action state so mouse-away doesn't lose
|
||||
// the value.
|
||||
let purgeModal = $state<{ days: number } | null>(null);
|
||||
function openPurge() {
|
||||
purgeModal = { days: 30 };
|
||||
}
|
||||
function closePurge() {
|
||||
purgeModal = null;
|
||||
}
|
||||
async function confirmPurge() {
|
||||
if (!purgeModal) return;
|
||||
const days = Math.max(1, Math.floor(purgeModal.days));
|
||||
markBusy('purge', true);
|
||||
try {
|
||||
const res = await purgeJobRuns(days);
|
||||
ui.notify(
|
||||
t(
|
||||
'admin.jobs.purge_done',
|
||||
{ n: res.purged, days: res.retention_days },
|
||||
'{{n}} old run(s) purged (retention {{days}} days)'
|
||||
),
|
||||
'success'
|
||||
);
|
||||
purgeModal = null;
|
||||
await loadJobs();
|
||||
} catch (e) {
|
||||
ui.notify(errorMessage(e), 'error');
|
||||
} finally {
|
||||
markBusy('purge', false);
|
||||
}
|
||||
}
|
||||
|
||||
// ─── Loading + polling ─────────────────────────────────────────────
|
||||
|
||||
/**
|
||||
@@ -459,8 +495,8 @@
|
||||
)}
|
||||
</p>
|
||||
</div>
|
||||
{#if hasBatch}
|
||||
<div class="jobs-panel__header-actions">
|
||||
<div class="jobs-panel__header-actions">
|
||||
{#if hasBatch}
|
||||
<button
|
||||
class="jobs-panel__btn jobs-panel__btn--primary"
|
||||
disabled={busyKeys.has('trigger:consistency_batch')}
|
||||
@@ -481,8 +517,25 @@
|
||||
<Icon name="play" />
|
||||
{t('admin.jobs.run_deep', 'Run deep')}
|
||||
</button>
|
||||
</div>
|
||||
{/if}
|
||||
{/if}
|
||||
<!-- Purge is orthogonal to consistency — it works even
|
||||
when the batch coordinator isn't registered, so it
|
||||
lives outside the {#if hasBatch}. Opens a modal so
|
||||
the operator picks a retention window with intent
|
||||
(no accidental delete-all). -->
|
||||
<button
|
||||
class="jobs-panel__btn"
|
||||
title={t(
|
||||
'admin.jobs.purge_hint',
|
||||
'Delete completed and failed run history older than the chosen retention window. Findings drop with their parent runs. Non-terminal runs are always preserved.'
|
||||
)}
|
||||
onclick={openPurge}
|
||||
disabled={busyKeys.has('purge')}
|
||||
>
|
||||
<Icon name="trash-alt" />
|
||||
{t('admin.jobs.purge', 'Purge old runs')}
|
||||
</button>
|
||||
</div>
|
||||
</header>
|
||||
|
||||
{#if loadError}
|
||||
@@ -889,6 +942,58 @@
|
||||
{/if}
|
||||
</section>
|
||||
|
||||
<!-- Purge modal — pick a retention window, then confirm. -->
|
||||
<Modal
|
||||
open={purgeModal !== null}
|
||||
title={t('admin.jobs.purge_title', 'Purge old job history')}
|
||||
onclose={closePurge}
|
||||
>
|
||||
{#if purgeModal}
|
||||
<form
|
||||
class="jobs-panel__purge-form"
|
||||
onsubmit={(e) => {
|
||||
e.preventDefault();
|
||||
void confirmPurge();
|
||||
}}
|
||||
>
|
||||
<p>
|
||||
{t(
|
||||
'admin.jobs.purge_body',
|
||||
'Delete completed and failed run history older than the chosen number of days. Findings drop with their parent runs. Non-terminal runs (running, paused, cancel-requested) are always preserved.'
|
||||
)}
|
||||
</p>
|
||||
<label>
|
||||
<span>{t('admin.jobs.purge_days_label', 'Retention (days)')}</span>
|
||||
<input
|
||||
type="number"
|
||||
min="1"
|
||||
step="1"
|
||||
bind:value={purgeModal.days}
|
||||
disabled={busyKeys.has('purge')}
|
||||
/>
|
||||
</label>
|
||||
<div class="jobs-panel__purge-actions">
|
||||
<button
|
||||
type="button"
|
||||
class="jobs-panel__btn"
|
||||
onclick={closePurge}
|
||||
disabled={busyKeys.has('purge')}
|
||||
>
|
||||
{t('common.cancel', 'Cancel')}
|
||||
</button>
|
||||
<button
|
||||
type="submit"
|
||||
class="jobs-panel__btn jobs-panel__btn--danger"
|
||||
disabled={busyKeys.has('purge')}
|
||||
>
|
||||
<Icon name="trash-alt" />
|
||||
{t('admin.jobs.purge_confirm', 'Purge')}
|
||||
</button>
|
||||
</div>
|
||||
</form>
|
||||
{/if}
|
||||
</Modal>
|
||||
|
||||
<style>
|
||||
.jobs-panel {
|
||||
display: flex;
|
||||
@@ -1192,6 +1297,28 @@
|
||||
flex-wrap: wrap;
|
||||
}
|
||||
|
||||
.jobs-panel__purge-form {
|
||||
display: flex;
|
||||
flex-direction: column;
|
||||
gap: 0.75rem;
|
||||
}
|
||||
|
||||
.jobs-panel__purge-form label {
|
||||
display: flex;
|
||||
flex-direction: column;
|
||||
gap: 0.25rem;
|
||||
}
|
||||
|
||||
.jobs-panel__purge-form input {
|
||||
max-width: 8rem;
|
||||
}
|
||||
|
||||
.jobs-panel__purge-actions {
|
||||
display: flex;
|
||||
gap: 0.5rem;
|
||||
justify-content: flex-end;
|
||||
}
|
||||
|
||||
.jobs-panel__link {
|
||||
display: inline-flex;
|
||||
align-items: center;
|
||||
|
||||
@@ -1241,6 +1241,13 @@
|
||||
"n_findings": "{{n}} findings",
|
||||
"n_notices": "{{n}} notices",
|
||||
"notices_present_tooltip": "Informational findings — no action required. Expand for detail.",
|
||||
"purge": "Purge old runs",
|
||||
"purge_hint": "Delete completed and failed run history older than the chosen retention window. Findings drop with their parent runs. Non-terminal runs are always preserved.",
|
||||
"purge_title": "Purge old job history",
|
||||
"purge_body": "Delete completed and failed run history older than the chosen number of days. Findings drop with their parent runs. Non-terminal runs (running, paused, cancel-requested) are always preserved.",
|
||||
"purge_days_label": "Retention (days)",
|
||||
"purge_confirm": "Purge",
|
||||
"purge_done": "{{n}} old run(s) purged (retention {{days}} days)",
|
||||
"state_running": "running",
|
||||
"never": "never",
|
||||
"just_now": "just now",
|
||||
|
||||
@@ -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() {
|
||||
|
||||
@@ -164,6 +164,9 @@ pub fn admin_routes() -> Router<Arc<AppState>> {
|
||||
"/jobs/{name}/runs/{id}/findings",
|
||||
get(list_job_run_findings),
|
||||
)
|
||||
// Retention cleanup — operator-triggered, not periodic.
|
||||
// See `purge_job_runs` docstring for the semantics.
|
||||
.route("/jobs/runs/purge", post(purge_job_runs))
|
||||
// Drives — admin-wide view (distinct from `/api/drives` which
|
||||
// is filtered to the caller's role grants).
|
||||
.route("/drives", get(list_all_drives))
|
||||
@@ -2392,3 +2395,74 @@ pub async fn list_job_run_findings(
|
||||
Err(e) => AppError::internal_error(format!("list_findings failed: {e}")).into_response(),
|
||||
}
|
||||
}
|
||||
|
||||
/// Query parameters for `POST /api/admin/jobs/runs/purge`.
|
||||
///
|
||||
/// `days` — retention window. Terminal runs (`Completed`, `Failed`)
|
||||
/// with `completed_at` older than this many days ago are deleted
|
||||
/// (with their findings via CASCADE). Default 30. Minimum enforced
|
||||
/// at 1 by the provider — zero would eat runs completed seconds
|
||||
/// ago. Non-terminal runs are ALWAYS preserved regardless of age.
|
||||
#[derive(serde::Deserialize)]
|
||||
pub struct PurgeJobRunsQuery {
|
||||
#[serde(default = "default_purge_days")]
|
||||
pub days: i32,
|
||||
}
|
||||
|
||||
fn default_purge_days() -> i32 {
|
||||
30
|
||||
}
|
||||
|
||||
/// `POST /api/admin/jobs/runs/purge?days=N` — operator-triggered
|
||||
/// cleanup of old terminal runs + their findings. Not periodic;
|
||||
/// admins fire this when they want to reclaim `jobs.*` history
|
||||
/// space. Delegates entirely to
|
||||
/// `JobStoreProvider::purge_terminal_runs` — no SQL in the handler
|
||||
/// (see `AGENTS.md` § handler thinness).
|
||||
#[utoipa::path(
|
||||
post,
|
||||
path = "/api/admin/jobs/runs/purge",
|
||||
params(
|
||||
("days" = Option<i32>, Query, description = "Retention window in days (default 30, minimum 1). Terminal runs older than this are deleted with their findings; non-terminal runs are always preserved."),
|
||||
),
|
||||
responses(
|
||||
(status = 200, description = "Purge complete"),
|
||||
(status = 401, description = "Unauthorized"),
|
||||
(status = 403, description = "Admin required"),
|
||||
(status = 500, description = "DB error"),
|
||||
),
|
||||
security(("bearerAuth" = [])),
|
||||
tag = "admin"
|
||||
)]
|
||||
pub async fn purge_job_runs(
|
||||
State(state): State<Arc<AppState>>,
|
||||
axum::extract::Query(query): axum::extract::Query<PurgeJobRunsQuery>,
|
||||
) -> impl IntoResponse {
|
||||
use crate::infrastructure::scheduler::JobStoreProvider as _;
|
||||
let retention_days = query.days.max(1);
|
||||
match state
|
||||
.core
|
||||
.job_store_provider
|
||||
.purge_terminal_runs(retention_days)
|
||||
.await
|
||||
{
|
||||
Ok(purged) => {
|
||||
tracing::info!(
|
||||
target: "audit",
|
||||
event = "jobs.runs_purged",
|
||||
purged = purged,
|
||||
retention_days = retention_days,
|
||||
"👮🏻♂️ admin purged {purged} terminal recoverable-run row(s) past {retention_days} day retention (findings cascaded)",
|
||||
);
|
||||
(
|
||||
StatusCode::OK,
|
||||
Json(serde_json::json!({
|
||||
"purged": purged,
|
||||
"retention_days": retention_days,
|
||||
})),
|
||||
)
|
||||
.into_response()
|
||||
}
|
||||
Err(e) => AppError::internal_error(format!("purge failed: {e}")).into_response(),
|
||||
}
|
||||
}
|
||||
|
||||
Reference in New Issue
Block a user