Queue backlog
Jobs of julia-pr and julia-ci builds that are ready to run but
have no agent yet, per pool (queue, OS and architecture), from
the jobs' Buildkite timestamps. Longest wait is how long the
oldest of them has been waiting; jobs waiting per slot divides
the count by the agent slots seen for the pool in the last 30
days. Where a pool's hosts also run another pool's jobs (the
Macs and the Windows and FreeBSD VM hosts), the tooltip says
so: those slots are only free while the other pool is idle. The second menu limits the chart to master or pull
request builds. The tooltip splits each pool's jobs between
them and adds the jobs running. Jobs are kept for 60
days.
Connected agents
Snapshots of the Buildkite agents API taken by every update
run, all queues and pipelines. The build, test, launch and
default queues have no resident agents: each host starts one
agent per job and it disconnects when the job ends, so they are
listed per host with the slots seen mid-job, and a host is
flagged once no slot of it has run a job for 3 days. Any other
agent seen in the last week but not in the latest snapshot is
flagged as missing.
Loading agent snapshots...
Loading worker history...
Each point is one master build: its wall time
from the first job starting to the last finishing, the median
wait of its jobs between becoming runnable and starting on an
agent (the queue), and the total job time it consumed. Builds
fetched from the Buildkite API since the database started carry
the timestamps; click a point or row to open the build.
Loading...