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
Even if N is large it can be inaccurate for the reasons Mark describes. Another example is pushing a template update that dramatically affects build time. In that case it could take two weeks for the bar to return to accuracy. This problem was my largest hesitation to building this feature. Some alternatives:
We can take the coefficient of variation across the builds, and if it's above a threshold not show a progress bar
We can ditch the progress bar and just show an ETA like "1m to 3m remaining..." where "1m" is the 5th percentile and 3m is the 95th percentile
I think progress is probably best if it's not measured in time because of the reasons you've outlined. Some ideas of things that are relatively static despite the changes you've identified:
Static milestones in the job like queued, downloaded, running tf, etc (most basic version)
Number of lines of output, the average lines of output should be more stable despite template changes, etc.
Either way I don' think we need so be accurate here so much as we just need to show them something is moving along.
Activity
kylecarbs commentedon Oct 21, 2022
@ammario we need to hide the progress bar unless there are N amount of builds to collect data from. Right now, it shows on the first workspace build.
ammario commentedon Oct 21, 2022
Even if N is large it can be inaccurate for the reasons Mark describes. Another example is pushing a template update that dramatically affects build time. In that case it could take two weeks for the bar to return to accuracy. This problem was my largest hesitation to building this feature. Some alternatives:
[-]build running progress bar is inaccurate [/-][+]progress bar: build running is inaccurate [/+]f0ssel commentedon Oct 31, 2022
I think progress is probably best if it's not measured in time because of the reasons you've outlined. Some ideas of things that are relatively static despite the changes you've identified:
Either way I don' think we need so be accurate here so much as we just need to show them something is moving along.