Queues
qwait: and qbacklog: — per-queue health. This had its own page until 6.0; the data was always these
same two folds, so it now sits beside every other durable metric.
| Column | Means |
|---|---|
| Count | Jobs claimed — the queue-wait sample count |
| Backlog | Eligible jobs waiting right now |
| Oldest | Age of the frontmost waiting job |
| Avg / p95 / p99 | Time between a job becoming eligible and a worker claiming it |
Two things make this tab unlike the others. It is the only family reporting two percentiles, because a queue is judged on its tail rather than its middle. And Backlog / Oldest are a gauge sampled every ~60s rather than a fold — which is why the family names its count token explicitly, so counting the gauge as an observation cannot deflate the average wait.
A rising backlog with a growing oldest-age is the classic "not enough worker capacity for this queue" signal. Wait is measured from the moment a job became eligible, so a requeue restarts the clock.
See Queue metrics for how the numbers are produced.