1423446 - [meta] Remove content process from probes only recorded in parent

Status: NEW

(Toolkit :: Telemetry, enhancement, P3)

Description

[Frank Bertsch [:frank]]

Running some analysis, I found that for 20171201, not a single ping reported SUBPROCESS_CRASHES_WITH_DUMP for the content process [0]. :chutten confirmed that it is chrome-only.

Can we remove content for other crash-related probes as well?

[0a] https://screenshots.firefox.com/31LIVpX0pRSsQC6O/dbc-caf9527b-e073.cloud.databricks.com
[0b - Databricks] https://dbc-caf9527b-e073.cloud.databricks.com/#notebook/3829/command/3833

Comment 1

[Chris H-C :chutten]

We can likely remove it from _many_ probes. ["main", "content"] was a default value populated to nearly all probes when it became a required field in bug 1335343.

Anything that runs in ipc/glue/CrashReporterHost.cpp is a good candidate for ["main"]. Ditto anything recorded from ContentParent, most jsms, and and and...

Comment 2

[Frank Bertsch [:frank]]

Attachment: aggregates-parent-process-probes.csv

In the interest of looking into this a bit more, I found all the probes that only reported the parent process from submission dates 20171201-20171205, running nightly 57 from the aggregates dataset.

Note that this includes many keyed histograms (each key is a row), as well as simple measures.

Comment 3

[Frank Bertsch [:frank]]

Attachment: parent-process-probes.sql

For posterity, here is the query I ran. We can easily look at more dates, builds, or versions.

Comment 4

[Chris H-C :chutten]

Simple measurements have no registry, but they are reported on childPayloads. I'd expect there to be data from those, unless the aggregator's never done it and I just haven't noticed.

Comment 5

[Frank Bertsch [:frank]]

(In reply to Chris H-C :chutten from comment #4)
> Simple measurements have no registry, but they are reported on
> childPayloads. I'd expect there to be data from those, unless the
> aggregator's never done it and I just haven't noticed.

There are some that are not present there, e.g. totalTime, uptime, and UITelemetry (e.g. these must have had non-parent processes, or they just weren't reported at all during the period?)

Not to worry, we do aggregate content simple_measures: https://github.com/mozilla/python_mozaggregator/blob/master/mozaggregator/aggregator.py#L298

Comment 6

[Georg Fritzsche [:gfritzsche]]

We can break this out into a few mentored bugs.

Comment 7

[Swapnil P (:DarthSwap)]

Hello, I am interested in working on this. Can you give me some pointers on where to start?

Comment 8

[Chris H-C :chutten]

This is a meta bug: a bug that coordinates other bugs. The actual work will happen in other bugs from within the "Depends on" list. I have just filed a new one, bug 1430603, that you are welcome to volunteer for, if you'd like.

Comment 9

[William Lachance (:wlach)]

This was done with some of the tab spinner probes in bug 1476071