1438258 - Track (at least) crashes in fennec

Status: RESOLVED FIXED

(Data Platform and Tools Graveyard :: Datasets: Error Aggregates, enhancement, P1)

Description

[William Lachance (:wlach)]

Now that bug 1407557 is implemented, we should at least start tracking fennec crashes within the error_aggregates. This is a prerequisite for monitoring fennec within missioncontrol.

Comment 1

[Mark Reid [:mreid]]

Frank, can we use the "durations" field from the core ping to get a measure of kilo-usage-hours?

Per the core ping docs[1], it is "combined duration, in seconds, of all sessions since last upload".


[1] https://firefox-source-docs.mozilla.org/toolkit/components/telemetry/telemetry/data/core-ping.html

Comment 2

[Frank Bertsch [:frank]]

Yup, the durations field is the way to go here. This will dovetail nicely with some of the Amplitude work where I also need to parse the core ping.

Some of the dimensions and metrics don't match up. For example, is gfx_compositor reported correctly (is that even a thing on mobile?).

None of the histograms will be available - neither the count histograms or the thresholds, nor will first_paint.

Comment 3

[William Lachance (:wlach)]

(In reply to Frank Bertsch [:frank] from comment #2)
> Yup, the durations field is the way to go here. This will dovetail nicely
> with some of the Amplitude work where I also need to parse the core ping.
> 
> Some of the dimensions and metrics don't match up. For example, is
> gfx_compositor reported correctly (is that even a thing on mobile?).

Fairly certain it isn't.
 
> None of the histograms will be available - neither the count histograms or
> the thresholds, nor will first_paint.

I think that's all ok, we can just return null or 0 where we don't have a value for a dimension or count.

Comment 4

[Marcia Knous [:marcia]]

Hello Will - Do we have any idea when this work will be started? Thanks!

Comment 5

[William Lachance (:wlach)]

I asked :mreid about this yesterday, I think he knows what's going on.

Comment 6

[Mark Reid [:mreid]]

Quick update here.

It looks like we're getting close to unblocking deployment of the latest telemetry-streaming code. That should happen next week, barring new problems coming to light.

Once that has happened, we will need to make a change to the Error Aggregates code to:
- Include crash pings where application = Fennec
- Process core pings where application = Fennec

This will give us crash counts and usage hours.

Frank, you mentioned that it can be tricky to add a new ping type, could you set the points on this bug based on how much effort you expect the above changes to Error Aggregates to be?

Comment 7

[Frank Bertsch [:frank]]

This should be doable with 3 points of effort.

Just to clarify, we will be processing:

- Crash pings with appName == "Fennec"
- Core pings with appName == "Fennec" AND os == "Android"

Comment 8

[William Lachance (:wlach)]

Note that crash counts on beta won't have the correct version until bug 1453238 is fixed.

Comment 9

[GitHub Bugzilla PR Linker]

Attachment: Link to GitHub pull-request: https://github.com/mozilla/telemetry-streaming/pull/129