1065880 - measure use of e10s

Status: RESOLVED FIXED

(Firefox :: General, defect)

Description

[:Gavin Sharp [email: [e-mail]]]

I want to know:
- what percentage of our Nightly population has browser.remote.tabs.autostart set to true (i.e. are testing "full e10s")
- what percentage of our Nightly population has an e10s window open (i.e. are testing "partial e10s")

This would be particularly useful to measure the effects of bug 1064885 and bug 1064886.

Do we have any ways to answer that question yet? Couldn't see anything relevant in the FHR docs, and I don't see any relevant telemetry either. Can we get something quickly?

Comment 1

[Benjamin Smedberg]

Indeed, I don't see this measurement in telemetry or FHR. For now I think we can have a good-enough solution for this using telemetry probes (measuring sessions instead of users). Somebody on e10s team ought to be implement, but gfritzsche or I can review.

Comment 2

[:Felipe Gomes (needinfo for replies!)]

Attachment: E10s telemetry

Comment 3

[:Gavin Sharp [email: [e-mail]]]

Additional questions whose answers would be interesting:
- how many sessions have browser.enabledE10SFromPrompt=true and browser.tabs.remote.autostart=true?
- how many sessions have browser.enabledE10SFromPrompt=true and browser.tabs.remote.autostart=false?

Comment 4

[Benjamin Smedberg]

Comment on attachment 8488016 [details] [diff] [review]
E10s telemetry

This answers the original question. Happy to take another patch to deal with browser.enabledE10SFromPrompt measurement.

Comment 5

[Chris Peterson [:cpeterson]]

Felipe, is your r+'d patch ready for checkin?

Comment 6

[:Felipe Gomes (needinfo for replies!)]

It's ready, but I'm hoping to get the follow-up in today too. If the followup is not ready for today I'll land this patch later in the afternoon.

Comment 7

[:Felipe Gomes (needinfo for replies!)]

Attachment: Measure prompt turnout

Benjamin, I think this is what Gavin wants to measure in comment 3. What do you think?

Comment 8

[:Felipe Gomes (needinfo for replies!)]

Attachment: Measure blacklisted e10s

And another interesting measurement that I want to throw out there if someone else agrees it is useful: number of sessions that were blocked from running e10s because safe-mode/a11y/ime was active.

Comment 9

[Mike Conley (:mconley) (:⚙️)]

Comment on attachment 8488759 [details] [diff] [review]
Measure blacklisted e10s

Review of attachment 8488759 [details] [diff] [review]:
-----------------------------------------------------------------

::: toolkit/xre/nsAppRunner.cpp
@@ +4525,5 @@
>  {
>    if (!gBrowserTabsRemoteAutostartInitialized) {
> +    bool prefEnabled = Preferences::GetBool("browser.tabs.remote.autostart", false) ||
> +                       Preferences::GetBool("browser.tabs.remote.autostart.1", false);
> +    gBrowserTabsRemoteAutostart = !gSafeMode && prefEnabled;

Ah, yes, that looks much better anyways.

Comment 10

[:Felipe Gomes (needinfo for replies!)]

https://hg.mozilla.org/integration/mozilla-inbound/rev/fa0fe907570c

Comment 11

[:Felipe Gomes (needinfo for replies!)]

Follow up fix: https://hg.mozilla.org/integration/mozilla-inbound/rev/538eb2f60373

Comment 12

[Phil Ringnalda (:philor)]

https://hg.mozilla.org/mozilla-central/rev/fa0fe907570c
https://hg.mozilla.org/mozilla-central/rev/538eb2f60373

Comment 13

[[SV Manager] Florin Mezei, QA (:FlorinMezei)]

This has been implemented almost 2 months ago in Nightly. I'm guessing that if there were some issues it would have been reported by now, so setting this as qe-verify- since we have other higher priority work.