1128003 - Experiments: ensure that experiments can withstand manifest changes in sample, endTime, maxRunningSeconds

Status: RESOLVED WONTFIX

(Firefox Health Report Graveyard :: Client: Desktop, defect, P4)

Description

[:Felipe Gomes (needinfo for replies!)]

It will probably come in handy in the future the ability to change the sampling size of a running experiment (to release it gradually, or increase or reduce sampling as necessary, according to data coming in), and to extend or shorten the length of an experiment by changing the endTime or maxRunningSeconds.

In one of the earlier experiments I suggested starting with a small sample size and changing it later, but as it was something that wasn't tried before, we didn't want to do it live to figure out.

It may well be that all of this works, but it would be nice to have existing evidence of that (and test coverage) to ensure we can do this with peace of mind whenever it becomes necessary.

Comment 1

[Georg Fritzsche [:gfritzsche]]

They *should* get re-evaluated properly.
We can rather easily add test-coverage here:
browser/experiments/test/xpcshell/test_api.js

Comment 2

[Georg Fritzsche [:gfritzsche]]

Telemetry experiments were removed from Firefox per bug 1415284.