Closed
Bug 1438564
Opened 8 years ago
Closed 8 years ago
Cannot access to dual stack DNS sites if IPv6 breaks while the browser being kept open
Categories
(Core :: Networking, defect, P2)
Tracking
()
RESOLVED
DUPLICATE
of bug 1430659
People
(Reporter: timdream, Assigned: mayhemer)
Details
(Whiteboard: [necko-triaged])
Attachments
(1 file)
|
2.31 MB,
application/zip
|
Details |
STR:
1. On a home wifi network with both IPv4/v6 enabled (and working), access https://mail.google.com/ and leave the tab open.
2. Wait until the IPv6 service breaks down (I've not figure out the cause, but |traceroute6 www.google.com| stopped at the router.)
3a. Inspect the opened GMail tab.
3b. Attempt to open www.google.com in a new tab
3c. Fire up Nightly with another profile, attempt to access www.google.com
Expected:
3a. GMail should continue to work since IPv4 still works (and Happy Eyeball should kick in?)
3b. www.google.com should continue to work too
3c. www.google.com should work
Actual:
3a. GMail shows the yellow "Unable to connect to network" banner.
3b. cannot connect to www.google.com with request timeout. DevTool Network Monitor says DNS timeout, after 70xxx ms.
3c. www.google.com connects correctly without delay via IPv4.
Note:
1. I can reproducible this bug every time (2) happens but I don't have control over when it happens. So I am not sure if I could do any log capture.
2. The fact Firefox didn't fallback to IPv4 for case (3b) doesn't feel right to me, but feel free to wontfix if this is by design, and the user must restart the browser.
Tim: sounds like there could be a couple issues here. 3a may be some strange interaction between h2 persistent connections and the hosed ipv6 link (perhaps we're not noticing the dead connection for some reason?) Does this continue happening if you disable http/2 (network.http.spdy.enabled.http2 = false)?
3b is indeed strange, though as you say perhaps as designed. I believe Honza did a lot of work on our happy eyeballs stuff, so I'll ni? him as well to see if he has any insight.
Flags: needinfo?(timdream)
| Reporter | ||
Comment 2•8 years ago
|
||
(Just to heads up, I won't be able to answer your question for another 3 weeks since I am away from the said setup physically.)
| Assignee | ||
Comment 3•8 years ago
|
||
We probably remember the host (all hosts involved) as ipv6 capable and from some reason never try ipv4 again.
Tim:
- I'll ask for a log before I start looking into this, it's fine if it's in 3 weeks, this is not a recent regression: https://developer.mozilla.org/en-US/docs/Mozilla/Debugging/HTTP_logging#Logging_HTTP_activity_by_manually_setting_environment_variables
- what exactly means "IPv6 service breaks down" for you? existing connections are killed? new conns can't be established?
- can I somehow simulate this locally?
Flags: needinfo?(honzab.moz)
| Reporter | ||
Comment 4•8 years ago
|
||
This is produced on Nightly. I started the log from about:networking, and reload www.google.com.tw again to wait for it to timeout.
Flags: needinfo?(timdream)
Putting your name on this Honza since the description resembles things you've poked on recently. Feel free to unassign if you think its wrong.
Assignee: nobody → honzab.moz
Priority: -- → P2
Whiteboard: [necko-triaged]
| Assignee | ||
Comment 6•8 years ago
|
||
Thanks for the log, a clear duplicate of bug 1430659
Status: NEW → RESOLVED
Closed: 8 years ago
Resolution: --- → DUPLICATE
You need to log in
before you can comment on or make changes to this bug.
Description
•