Changelog
Delivery latency
In short
Track how long each destination takes to respond
Outpost now records latency_ms on every delivery attempt: the time the destination took to respond or fail.
GET /metrics/attempts adds avg_latency, p50_latency, p95_latency and p99_latency measures. They work with the existing dimensions, filters and granularity, so you can chart response times per tenant or destination, or find the destinations that are slowing down. Failed attempts count too, so filter by status to look at successful deliveries only.
Attempts recorded before latency tracking return null. See metrics.
Thanks to @kunwar-axilio for contributing delivery latency tracking. It is available now in v1.5.0.