f703fba1ef
* fix: log cleanup scans S3 orphans and works cross-server Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com> * fix: don't skip service log orphan scan when job retention is disabled Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com> * fix: time-based heartbeat + flag partial folder sizes on list errors Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com> * refactor: move background_task module from common to api-settings Only log_cleanup and storage_usage use it today, both in windmill-api-settings. Keeping it in the consumer crate narrows the blast radius; if workers or indexer later need cross-server lease+progress coordination they can move it back to common then. Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com> --------- Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>
20 lines
894 B
SQL
20 lines
894 B
SQL
-- Tracks the live state of long-running background tasks coordinated across
|
|
-- multiple server replicas (log cleanup, object-storage usage scan, staggered
|
|
-- graceful restart). Previously these were stored in-memory per-process or in
|
|
-- `global_settings` (which is conceptually for user-configurable instance
|
|
-- settings and is exposed via the settings UI / export).
|
|
--
|
|
-- `value` holds task-specific JSON. `running`/`owner`/`updated_at` implement
|
|
-- a lease: a task is "free" if running=false OR updated_at is older than the
|
|
-- heartbeat-stale threshold (handled in application code, see
|
|
-- windmill-common/src/background_task.rs).
|
|
CREATE TABLE background_task_state (
|
|
name TEXT PRIMARY KEY,
|
|
value JSONB NOT NULL,
|
|
running BOOLEAN NOT NULL DEFAULT FALSE,
|
|
owner TEXT,
|
|
started_at TIMESTAMPTZ,
|
|
finished_at TIMESTAMPTZ,
|
|
updated_at TIMESTAMPTZ NOT NULL DEFAULT NOW()
|
|
);
|