1187054 - v4 totalTime not binned by subsession

Status: RESOLVED WORKSFORME

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

Description

[brendan c]

See the notebook here--
https://gist.github.com/bcolloran/323eeea11de94ca8b2dc
-- it appears that totalTime appears to increase monotonically from one subsession to the next within a client session (but activeTicks does not, it just looks crazy). This behavior is not correct; in both cases, the value recorded for activeTicks and totalTime should be the value for that subsession only, not the cumulative total for the entire client session.

The reason for this is that when a client session is split across days or environment change, we want to know the time spent in each day, or in each environment state without having to find a previous ping and subtract. The values collected in the subsession bins should be non-overlapping so that they can be aggregated without risk of double counting.

The same applies to e.g. search, crashes, basically anything else that happens in time-- it must be assigned to a subsession so that we can aggregate by environment, date, client, etc without double counting.

Comment 1

[brendan c]

Note that my first guess is that I'm looking at the wrong field. Is there a field that records the time per subsession? (If so, the v2/v4 data equivalents spreadsheet should be updated with that field).

Comment 2

[Alessio Placitelli [:Dexter]]

Yes, |totalTime| tracks the time spent in the whole session, is not reset on subsession changes.

We do have the |subsessionLength| field [1], which tracks the subsession length in seconds. Is this what you were looking for?

[1] - https://ci.mozilla.org/job/mozilla-central-docs/Tree_Documentation/toolkit/components/telemetry/telemetry/main-ping.html

Comment 3

[brendan c]

Thanks guys, that was the problem, very helpful. I added a note to the v2/v4 equivalency sheet.

Still some more things to look into, but on the right track now at least.

https://gist.github.com/bcolloran/f79914990e88da62eeaa

Comment 4

[Georg Fritzsche [:gfritzsche]]

Thought i'd close this one as the original question is resolved, lets file new bugs on other specific problems found?

Comment 5

[brendan c]

Sure, sounds good.