Closed Bug 1560068 Opened 6 years ago Closed 5 years ago

Discuss what we should do with "uncommitted time" in time spans

Categories

(Data Platform and Tools :: Glean: SDK, defect, P3)

Product:
Data Platform and Tools
Component:
Glean: SDK
Type:
defect

Tracking

(Not tracked)

Status:
RESOLVED WONTFIX

People

(Reporter: chutten, Unassigned)

References

Details

(Whiteboard: [telemetry:glean-rs:backlog])

Imagine a timespan that measures the amount of time the app is in the foreground, but is in the "metrics" ping and has a "ping" lifetime. Let's call it metrics_foreground. The session starts at 0345 and the timespan is started. 0400 rolls around and the "metrics" ping is sent.

What happens?

Well, today nothing happens. The timespan isn't "committed" (there's nothing in it because stopAndSet hasn't been called).

Since metrics_foreground is in the "metrics" store, should we take the uncommitted time and put it in the "metrics" ping ('restarting' the timespan to 0 at 0400)?

Or, since we wouldn't do this for a timing distribution like some pageload metric, should we continue doing nothing? Though this would mean we'd have no way of including in a ping any metrics we wanted to have a clean "edge" that starts and ends with a store's ping.

Is it worth to get data science feedback/review on any potential solution here?
(See also: figuring out a process to propose & review metrics or significant changes to them cross-functionally)

Priority: -- → P3
Whiteboard: [telemetry:glean-rs:m?][telemetry:mobilesdk:m?]
Whiteboard: [telemetry:glean-rs:m?][telemetry:mobilesdk:m?] → [telemetry:glean-rs:m?][telemetry:mobilesdk:m8]

Bring in data science on it ;)

Flags: needinfo?(alessio.placitelli)
Whiteboard: [telemetry:glean-rs:m?][telemetry:mobilesdk:m8] → [telemetry:glean-rs:m8][telemetry:mobilesdk:m8]

Let's defer this discussion to the next milestone. Leaving the ni? around.

Whiteboard: [telemetry:glean-rs:m8][telemetry:mobilesdk:m8] → [telemetry:glean-rs:m8][telemetry:mobilesdk:m9]
Whiteboard: [telemetry:glean-rs:m8][telemetry:mobilesdk:m9] → [telemetry:glean-rs:backlog]
Depends on: 1570386
Flags: needinfo?(alessio.placitelli)
Depends on: 1608137

Marissa,Mike

do you have any thought on how/if comment 0 should be handled?

Flags: needinfo?(mgorlick)
Flags: needinfo?(mdroettboom)

I don't know if I have a suggestion solution, but want to call out a related problem. Definitely interested in hearing :mgorlick's thoughts as this relates to time-related analysis.

Related to this, we have the problem of uncommitted time when the application shuts down. Right now, we lose that time completely, which IMHO, is even worse than having that time placed on a later ping (in the scenario from comment #0). Given all of the problems we know we have with "doing things on process shutdown" with Android, I don't know if there is any solution to that problem.

Flags: needinfo?(mdroettboom)

reassigning to Leif.

Flags: needinfo?(mgorlick) → needinfo?(loines)

Wrong ticket! :) back to me.

Flags: needinfo?(loines) → needinfo?(mgorlick)
Flags: needinfo?(mgorlick)
Status: NEW → RESOLVED
Closed: 5 years ago
Resolution: --- → WONTFIX
You need to log in before you can comment on or make changes to this bug.