drovr CI speed
No refresh has finished yet. The first refresh runs within the hour.
Train CI now
no data yet
Merge to prod now
no data yet
PR CI, full mode
no data yet
Runner time per train
no data yet
Train CI
No runs recorded yet.
one run median of 7 runs outlier, left out of the line. Runs off the scale sit on its top edge.
One dot is one successful CI run on a release train (a train/* pull request), measured from its first job starting to its last job finishing, so queue time before the first job is left out. Failed and cancelled runs are left out, and so are re-run attempts. A dot more than 3× above or below the median of its 7 neighbours is an outlier: it is drawn in grey and left out of the line. The line is the median of each run and the 3 runs either side.
Merge to prod
Merge-to-prod source not wired yet. CD fires releases outside GitHub Actions, so the ship workflow this chart reads records none of them.
one run median of 7 runs outlier, left out of the line. Runs off the scale sit on its top edge.
One dot is one release: from the train's merge commit landing on main to its exact readback in prod. Releases that did not succeed on the first attempt are left out, as are plans that stopped for a risk approval. Outliers and the line work as in the train chart.
PR CI, full mode
No runs recorded yet.
one run median of 7 runs outlier, left out of the line. Runs off the scale sit on its top edge.
One dot is one successful CI run on an ordinary pull request in full mode (the native MySQL suites ran), from its first job starting to its last job finishing. Light-mode runs and runs that reused a proof from an earlier run are left out. Before CI modes existed (mid-October), every run was full. Outliers and the line work as in the train chart.
Runner time per train run
No runs recorded yet.
one run median of 7 runs outlier, left out of the line. Runs off the scale sit on its top edge.
One dot is the runner time one successful train run used: the sum of every job's start-to-finish minutes, on the self-hosted pool and GitHub-hosted runners alike. It measures cost, not waiting. It comes from job-level data, which starts with the first train. Outliers and the line work as in the train chart.