Org pool — concurrent slots (24h) click an hour to update the table on the right · drag to select a range

Each bar is the average concurrent slots over that Pacific wall-clock hour. The current hour is partial — its value is the average over the elapsed minutes, refreshed every 15 min by the cron. For sub-hour live load, see the "Currently running & pending" panel on a reservation's detail page.

Wait & queue (24h)

Bars: jobs whose wait (submission → first slot) exceeded the chosen threshold. Lines: org-wide weighted p50 / p95 wait. >5s removes scheduling jitter; >10s flips the banner red.

queue threshold > 

Query runtime (24h)

vs typical: org-wide p95 runtime line over a shaded p25–p75 band from the same hour-of-day across the trailing 14 days. Spikes above the band = "queries slower than usual." % distribution: share of jobs in each runtime bucket. avg / p90 / p95: weighted percentiles.

view:

Top consumers — current (updates with the org-pool selection above)

What's new (last 48h, absent in prior 14d)

Pool utilization — daily

Pain heatmap — reservation × hour-of-day

metric:

Per-reservation max slots vs baseline (small multiples)

Long-running jobs (≥ 5m runtime, daily, stacked by reservation)