1382820 - Convert a11y_consumers and a11y_instantiated_flag to opt-out

Status: RESOLVED FIXED

(Toolkit :: Telemetry, defect, P3)

Comment 1

[Jim Mathies [:jimm]]

Attachment: patch

Comment 2

[Jim Mathies [:jimm]]

Comment on attachment 8888518 [details] [diff] [review]
patch

For data review. We'd like to collect accurate client counts out on release for these probes. The instantiated flag is a bool, no privacy issue there afaict. The consumers probe is an enum that detects injected dlls to identify specific clients, which is no different from the recent opt-out consumer scalar we landed.

Comment 3

[Benjamin Smedberg]

Comment on attachment 8888518 [details] [diff] [review]
patch

There are a few blocking issues and a bunch of nonblocking issues that should be followups:

Blocking issues:
* Both histograms need bug_numbers and alert_emails. alert_emails needs to have a real person who's responsible for the long-term value of the data (it may also have a list address).
* The description of A11Y_CONSUMERS mentions an enum but not it's name, but I can't figure out what the recorded values mean. If there is an existing in-tree enumeration, at least point to it (by name or DXR link), or if there's not, the description should include what each value means.

Non-blocking issues:
* for A11Y_INSTANTIATED_FLAG, flag histograms are basically deprecated now, and it would be better to use a boolean scalar
* both of these say record_in_processes main+content but I suspect that they are main-process-only and that would improve various aggregations

Comment 4

[Benjamin Smedberg]

Oh, and when you add bug_numbers and alert_emails you can remove these histograms from the whitelist of old histograms that don't follow the rules (histogram-whitelists.json)

Comment 5

[Benjamin Smedberg]

I forgot one non-blocking issue (can be a followup): typically opt-out metrics need to have a unit test to ensure that they still work. This is important because code refactorings fairly frequently break telemetry without people noticing. So I would strongly encourage you to try and get basic unit tests working for these metrics. I realize this might be difficult because our testing framework probably doesn't do much a11y.

Comment 6

[Jim Mathies [:jimm]]

Attachment: patch

- added bug#s (this bug) to the two probes I'm converting to opt-out
- added emails for all four probes
- removed entries from white lists
- filed two follow ups, converting to boolean scalars and adding unit tests

Comment 7

[Benjamin Smedberg]

Comment on attachment 8889893 [details] [diff] [review]
patch

data-r=me thanks!

Comment 8

[Jim Mathies [:jimm]]

Attachment: patch

- updating email address to something more generic

Comment 9

[Jim Mathies [:jimm]]

Comment on attachment 8889894 [details] [diff] [review]
patch

+ from irc

Comment 10

[Jim Mathies [:jimm]]

Attachment: patch

s/been used/has been used text in comments.

Comment 11

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/mozilla-inbound/rev/547d9dfbbf67
Convert a11y_consumers and a11y_instantiated_flag to opt-out. r=davidb, r=bsmedberg

Comment 12

[Carsten Book [:Tomcat]]

https://hg.mozilla.org/mozilla-central/rev/547d9dfbbf67

Comment 13

[Jim Mathies [:jimm]]

Comment on attachment 8889896 [details] [diff] [review]
patch

Approval Request Comment
[Feature/Bug causing the regression]:
converting existing a11y telemetry probes to opt out
[User impact if declined]:
we'd like tyo have this data out on release for 55 so we can make some decisions about 56/57 rollout.
[Is this code covered by automated tests?]:
telemetry is
[Has the fix been verified in Nightly?]:
changed probe to opt out
[Needs manual test from QE? If yes, steps to reproduce]: 
no
[List of other uplifts needed for the feature/fix]:
none
[Is the change risky?]:
no
[Why is the change risky/not risky?]:
existing well test probe converted to opt out on release
[String changes made/needed]:
none

Comment 14

[Julien Cristau [:jcristau]]

Comment on attachment 8889896 [details] [diff] [review]
patch

convert 2 a11y telemetry probes to opt-out, needed in 55 release.  should be in 55.0b13

Comment 15

[Ryan VanderMeulen [:RyanVM]]

https://hg.mozilla.org/releases/mozilla-beta/rev/e2dd499956f0

Comment 16

[Andrei Vaida [:avaida]]

(In reply to Jim Mathies [:jimm] from comment #13)
> [Is this code covered by automated tests?]:
> telemetry is
> [Has the fix been verified in Nightly?]:
> changed probe to opt out
> [Needs manual test from QE? If yes, steps to reproduce]: 
> no

Setting qe-verify- based on Jim's assessment on manual testing needs.