SETA finds the minimum set of jobs to run in order to catch all failures that our automation has found in the recent past on Firefox development repositories. There’s one main API that SETA offers consumers (e.g. the Gecko decision task) in order to show which jobs are consider low value (less likely to catch a regression). After a certain number of calls, the API will return all jobs that need to be run.
SETA creates job priorities for all jobs found in the runnable-jobs API for that repository. Initially all jobs will be treated as low value, however, once we run the test to analyze past failures we will mark certain jobs as high value (priority=1).
Also note that jobs from different platforms (linux64 vs win7) or different CI systems (Buildbot vs TaskCluster) will be treated the same (use the same job priority). In other words, a job priority can represent a multiple number of jobs.
Jobs that appear on Treeherder for the first time will be treated as a job with high priority for a couple of weeks since we don’t have historical data to determine how likely they’re to catch a code regression.
In order to find open bugs for SETA visit list of SETA bugs.
- This is the API that consumers like the Gecko decision task will use
- This API shows which jobs are defined for each project
- This API shows job failures that have been annotated with “fixed by commit”
Local set up¶
After you set up Treeherder, ssh (3 different tabs) into the provisioned VM and follow these steps:
yarn install –no-bin-links yarn start:local
- Test the different APIs:
- This one won’t work until https://bugzilla.mozilla.org/show_bug.cgi?id=1389123 is fixed