right now SETA takes data from many sources, one of those is windows XP. We decommissioned this about 10 weeks ago, but we still have 6+ weeks of windows xp failure data influencing the results of SETA. To make this accurate, I propose a hack to remove windowsXP* data from the bucket of data we analyze. playing with this locally, I see us going from 148 high value jobs to 160 high value jobs- this says that windows xp has a real effect.
Could we do this in a more general way, say by discarding any data which we haven't seen instances of in the last week or so?
oh, I like that idea! this is why I need others besides me hacking on the SETA algorithm. That would make a more complicated query- or...we could cross reference with runnable_jobs.json maybe?
Status: NEW → RESOLVED
Last Resolved: 10 days ago
Resolution: --- → WONTFIX
You need to log in before you can comment on or make changes to this bug.