Closed Bug 1882745 Opened 1 year ago Closed 11 months ago

Intermittent DNS resolution failures.

Categories

(Core :: Networking: DNS, defect, P2)

Product:
Core
Component:
Networking: DNS
Type:
defect

Tracking

()

Status:
RESOLVED INCOMPLETE

People

(Reporter: freddy, Unassigned)

References

(Blocks 1 open bug)

Details

(Whiteboard: [necko-triaged][necko-priority-new])

Everyone once in a while I get intermittent page load errors for new tabs.
The steps are roughly:

  • Open new tab
  • Type URL
  • Page fails to load "Hmm. We're having trouble finding that site." almost immediately.

Configuration: I am not using DoH.

This seems to be an isolated problem specific to a content process. Other tabs all work fine.

I have finally managed to capture a profile, which I am sharing as https://profiler.firefox.com/public/rb4p1sz56kca978t6w51hd7a826ar0gbhv0bdzg/calltree/?globalTrackOrder=01&thread=3&v=10.

From my read, this seems to be a timeout waiting for the socket thread when performing the DNS request. The tab we end up loading is about:neterror?e=dnsNotFound&u=(...)

Retrying a couple of minutes later, everything just loads. Again, other tabs work fine and resolving DNS from the command line also works. Please let me know what else I can do to help solve this.

(In reply to Frederik Braun [:freddy] from comment #0)

From my read, this seems to be a timeout waiting for the socket thread when performing the DNS request. The tab we end up loading is about:neterror?e=dnsNotFound&u=(...)

I'm not sure if that's true. What makes you think that?

Retrying a couple of minutes later, everything just loads. Again, other tabs work fine and resolving DNS from the command line also works. Please let me know what else I can do to help solve this.

I'm thinking that a DNS request failed for some reason, and we put it in the cache. Since we key DNS cache entries using the top level origin, one tab might see DNS failures for the same domain. If this happens in the future, could you use the Networking Setting for profiling? Log messages don't show up for DNS threads for example if we're not profiling those.

Blocks: dns
Severity: -- → S3
Priority: -- → P2
Whiteboard: [necko-triaged][necko-priority-new]
Flags: needinfo?(fbraun)

(In reply to Valentin Gosu [:valentin] (he/him) from comment #1)

(In reply to Frederik Braun [:freddy] from comment #0)

From my read, this seems to be a timeout waiting for the socket thread when performing the DNS request. The tab we end up loading is about:neterror?e=dnsNotFound&u=(...)

I'm not sure if that's true. What makes you think that?

At the 1.892s timestamp, I see a DOMContentLoaded for about:neterror?e=dnsNotFound&u=(...).

Retrying a couple of minutes later, everything just loads. Again, other tabs work fine and resolving DNS from the command line also works. Please let me know what else I can do to help solve this.

I'm thinking that a DNS request failed for some reason, and we put it in the cache. Since we key DNS cache entries using the top level origin, one tab might see DNS failures for the same domain. If this happens in the future, could you use the Networking Setting for profiling? Log messages don't show up for DNS threads for example if we're not profiling those.

I have a full profile locally, but I'm not comfortable sharing everything on the public internet.
I already see a network track for the youtube.com content process. What's missing and where can I look?

Flags: needinfo?(fbraun)

Freddy, if you're still seeing this could you make sure to save the profile and reopen the bug?
Thanks!

Status: NEW → RESOLVED
Closed: 11 months ago
Resolution: --- → INCOMPLETE
You need to log in before you can comment on or make changes to this bug.