1695717 - Unexpectedly long delay when searching via Google and HTTP/3 enabled

Status: VERIFIED FIXED

(Core :: Networking, defect, P2)

Description

[Asif Youssuff]

Basic information

I have noticed for a few days that when doing a search via the built in Google search, I sometimes get a long delay. I used to work around it by doing the same search again (command-k, enter), but I took a few profiles today.

Steps to Reproduce:

1. Do a search via Command-k (separate search toolbar) or typing directly into address bar and hitting enter

Expected Results:

Very fast results.

Actual Results:

Multiple seconds waiting. Slow enough that I started to capture profiles while it was happening.

---

More information

Profile URL: Two profiles from today:

• https://share.firefox.dev/3rcT7go
• https://share.firefox.dev/302vVWc

Basic systems configuration:

OS version: macOS 10.15.7 (19H524)

GPU model: Intel HD Graphics 6000 1536 MB

Number of cores: 1.6 GHz Dual-Core Intel Core i5

Amount of memory (RAM): 4GB

Thanks so much for your help.

Comment 1

[Asif Youssuff]

Attachment: about:support

Comment 2

[Olli Pettay [:smaug][[e-mail]](vacation until Nov 16)]

That looks like Google is slow at responding.

Any chance you have some issues with the network connection?

Comment 3

[Olli Pettay [:smaug][[e-mail]](vacation until Nov 16)]

Moving to Networking but adding [qf] so that this stays in performance team's triage list.

Comment 4

[Henrik Skupin [:whimboo][⌚️UTC+2]]

Sometimes I experience the same slowness with Google specific domains. See my latest comments on bug 1648101. Maybe this is somewhat related. My suspicion so far is the usage of DoH and here especially NextDNS as provider. Triggering a forced reload of the page while it's still waiting to be loaded mostly causes the page to load instantly.

Note that when such a situation happens both Chrome and Safari also load the page instantly. Then AFAIR it also works for Firefox.

Asif, are you using DoH, and do you see a similar behavior when forcing a reload at this time?

Comment 5

[Asif Youssuff]

Henrik,

I'm a little surprised, but I am not using DoH on this profile of Firefox.

For what it is worth, I am using these DNS servers on my router:

9.9.9.9
1.1.1.1

The forced reload works for me as well.

Olli,

I don't have issues with general network connections - at least not that I have noted. I have a gigabit connection and I am connected via WiFi on a router in the same room.

Comment 6

[Henrik Skupin [:whimboo][⌚️UTC+2]]

This might be a dupe of bug 1694035. But I'll keep this bug open for now given that there might be other problems as well. We can check again once bug 1694035 has been investigated / fixed.

Comment 7

[Asif Youssuff]

I noticed that my install is in a "HTTP3 on Nightly" study - could that have anything to do with it?

Comment 8

[Kershaw Chang [:kershaw]]

(In reply to Asif Youssuff from comment #7)

I noticed that my install is in a "HTTP3 on Nightly" study - could that have anything to do with it?

Dragana, could you take a look?
Thanks.

Comment 9

[Henrik Skupin [:whimboo][⌚️UTC+2]]

I actually also have the HTTP3 experiment turned on.

Comment 10

[Dragana Damjanovic [:dragana]]

Can someone make a http log while this is happening. Is it reproducible enough to make a log?

https://developer.mozilla.org/en-US/docs/Mozilla/Debugging/HTTP_logging

When did it started happening? I have landed some changes on Friday and I am wondering if this was happening before Friday as well?

Comment 11

[Dragana Damjanovic [:dragana]]

Also Henrik, maybe you can reproduce

Comment 12

[Henrik Skupin [:whimboo][⌚️UTC+2]]

At the moment it happens kinda rarely for me. So maybe Asif has a better chance to catch it.

For me it's already happening a while, but lets see what Asif will say.

Comment 13

[Asif Youssuff]

Attachment: log.txt-main.8154.moz_log.zip

I grabbed a log using about:networking#logging. There seems to be a lot of log data, but I saw the slowness when searching for "manager" in the built in Google search in Firefox.

I'm not sure when it started happening - I run Nightly on all of my machines but generally on Linux. I have only seen this on my Mac, which I recently resumed using. I do think it was happening before Friday, however.

Comment 14

[Henrik Skupin [:whimboo][⌚️UTC+2]]

Dragana do you have an update for us? I'm hitting this kinda frequently now each and every day. It makes it kinda annoying to have to wait sometimes minutes. Thanks.

Comment 15

[Dragana Damjanovic [:dragana]]

Attachment: Make sure to trigger writes after HTTP/3 session has been connected

Comment 16

[Dragana Damjanovic [:dragana]]

A try build is here:
https://treeherder.mozilla.org/jobs?repo=try&revision=1bba5a7b93fecd49174c033385620fc3a3d8a59d

Can you try it out:
The Linux build is here: https://firefox-ci-tc.services.mozilla.com/api/queue/v1/task/IsFs0491RfegNZDFcy_jWg/runs/0/artifacts/public/build/target.tar.bz2

The MAC build is here: https://firefox-ci-tc.services.mozilla.com/api/queue/v1/task/OtAfZSMATjeh5Ub48Zdsew/runs/0/artifacts/public/build/target.dmg

Comment 17

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/autoland/rev/9bd0407c7178
Make sure to trigger writes after HTTP/3  session has been connected r=necko-reviewers,kershaw

Comment 18

[Dragana Damjanovic [:dragana]]

The change will be in Nightly tomorrow as well.

Comment 19

[Bogdan Tara[:bogdan_tara | bogdant]]

https://hg.mozilla.org/mozilla-central/rev/9bd0407c7178

Comment 20

[Dragana Damjanovic [:dragana]]

Comment on attachment 9213090 [details]
Make sure to trigger writes after HTTP/3 session has been connected

Beta/Release Uplift Approval Request

• User impact if declined: Page load may stall for a long time
• Is this code covered by automated tests?: No
• Has the fix been verified in Nightly?: No
• Needs manual test from QE?: No
• If yes, steps to reproduce:
• List of other uplifts needed: None
• Risk to taking this patch: Low
• Why is the change risky/not risky? (and alternatives if risky): A function (ResumSending) is called always after a a HTTP/3 connection enters state CONNECTED. There is one corner case when it is not called and this patch fixes it.
• String changes made/needed:

Comment 21

[Ryan VanderMeulen [:RyanVM]]

Comment on attachment 9213090 [details]
Make sure to trigger writes after HTTP/3 session has been connected

Approved for 88.0b7.

Comment 22

[Ryan VanderMeulen [:RyanVM]]

https://hg.mozilla.org/releases/mozilla-beta/rev/d9926575a6a2

Comment 23

[Henrik Skupin [:whimboo][⌚️UTC+2]]

Thanks Dragana for the fix! I will keep an eye on it over the next days if the page loads have been improved.

Comment 24

[Henrik Skupin [:whimboo][⌚️UTC+2]]

So far I haven't seen any issues anymore when HTTP/3 is enabled.