Tags

Tags give the ability to mark specific points in history as being important
  • 0.8.0

    Release 0.8.0: completed-jobs cursor on started_at time slices (scheduled jobs no longer missed, paging with a page cap and a 3s DB safety valve, overlap and dedupe, no double counting on restart, backoff for both cursors); new successful_jobs poller; schedule freshness checked per cron tick with a narrow query, stale no longer firing right after every tick; windmill_job_success_total and failure rate per script (panel and alert); preview-kind flow steps counted; new config FAILED_JOBS_OVERLAP, SUCCESSFUL_JOBS_OVERLAP, POLL_DELAY_SUCCESSFUL_JOBS, LIST_PAGE_SIZE, LIST_MAX_PAGES, LIST_MAX_PAGE_LATENCY, MAX_SUCCESSFUL_JOBS_WINDOW; cursor metrics renamed with a poller label.
  • 0.7.0

    Release 0.7.0: windmill_job_queue_wait_seconds (queue wait of every completed on-demand job, per priority tier) with on-time % and p95 wait panels and per-tier lateness alerts; README section on priority tiers; dashboard variables default to All; fake Windmill scenarios for late starts.
  • 0.6.1

    Release 0.6.1: config env vars (POLL_DELAY_*, DISCOVERY_INTERVAL, SCHEDULE_GRACE_PERIOD, MAX_FAILED_JOBS_WINDOW, LISTEN_ADDR, LOG_FORMAT, LOG_LEVEL) are actually read from the process environment — 0.5.0/0.6.0 silently ignored them; readable durations in the startup log.
  • 0.6.0

    Release 0.6.0: priority_tier values renamed to batch / async / interactive (was default / low / high), named after how soon a job must start (24h / 5m / ~10s, recommended Priority 0 / 25 / 75); per-tier queue alerts; failed-jobs alerts and panels count a script's first failure; Scripts dashboard: integer failure counts per interval, failures by folder with per-folder detail, personal folders toggle, collapsible Readme with priority rules.
  • 0.5.0

    Release 0.5.0: every flag configurable via env var (POLL_DELAY_*, DISCOVERY_INTERVAL, LOG_LEVEL, ...) with startup validation; windmill_exporter_poll_interval_seconds metric; suggested alerts relative to the configured interval; infra dashboard: container resources row (cAdvisor), poller delay in intervals; dashboard titles carry the exporter version.
  • 0.4.1

    Release 0.4.1: schedule timezones honored (IANA tz database embedded via time/tzdata; before, every non-UTC schedule fell back to UTC in the scratch image, skewing windmill_schedule_stale by 1-2h). Dashboards split into Infrastruttura and Script (sviluppatori), cursor-age/clamped-window panels, actionable panel descriptions, CODEOWNERS.
  • 0.4.0

    Release 0.4.0: single-name multi-arch image (windmill-exporter:<version> for amd64+arm64, replaces <version>-x86/-arm), schedule_health counts successes with null completed_at, dashboard rework (per-environment aggregations, increase-based failure panels, p90 queue wait), local docker-compose stack.
  • 0.3.0

    Release 0.3.0: failed_jobs poller hardening after the 2026-09-23 incident — per-workspace exponential backoff on poll errors, -max-failed-jobs-window cap on the started_after window (+ window_clamped_total counter), failed_jobs_cursor_age_seconds gauge, series cleanup on workspace departure.
  • 0.2.0

    Release 0.2.0: schedule-health poller (cron muto), EE dashboard panels, WINDMILL_TOKEN_FILE support for Swarm secrets, worker-metrics from v0.1.0 predecessor work.
  • v0.1.0

    Prima versione: exporter Prometheus per Windmill (failed-jobs + queue-age pollers, Docker/CI, docs)