You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
Nothing reports how fast the production lane is moving or how long it waits for runners #6206
The numbers needed are public job metadata and cost a few dozen API calls for a month of runs.
Problem
Runner starvation looks like a slow queue. Without a number nobody can tell the two apart, and nobody can say whether a fix worked.
Smallest useful change
One command that prints, for a chosen window: merges, time from queueing to merging, time per merge while backlogged, runner wait on the two jobs between production deploys, deploy duration, and evictions with how many later merged unchanged. It exits non-zero when the runner wait is above the target in #6203, so an agent run or a person can use it as a health check.
Activity