712741 - network.http.max-connections still causes Firefox to stop loading any page

Status: RESOLVED INCOMPLETE

(Core :: Networking: HTTP, defect)

Description

[Marco Bonardo [:mak]]

So, I can't confirm 100% the title of the bug is correct, but the issue I'm seeing is pretty much the same as in bug 692260, that is hard to verify fixed.

Randomly (but especially when some remote network like mozilla's ones is down), my browser stops loading any page, it sends out the get request and then hangs in connecting status. I reproduced this anytime some mozilla's networks were down (happened frequently in the last weeks). Edmorley reproduced something really similar.

Opening a new instance of the browser works fine, any other browser or software works fine, just this specific instance is hosed.

Now, looking at bug 692260, someone suggested an issue of the network card or router. The interesting fact is that I tried disabling and re-enabling the network cark in Windows devices management, and Firefox started loading pages again.

Comment 1

[Marco Bonardo [:mak]]

just to add some info my network chip is an RTL8111E, not that old.

Comment 2

[Marco Bonardo [:mak]]

note, the other times I saw the bug, I just restarted Firefox, that fixed the problem too.

Comment 4

[Marco Bonardo [:mak]]

Mossop points out maybe internal connections to AMO (and sync?) may have filled up the limit of connections, not allowing to load more pages.
If that's a possibility, I wonder how many connections do we open, and what's the timeout for those.

Comment 5

[Patrick McManus [:mcmanus]]

(In reply to Marco Bonardo [:mak] from comment #4)
> Mossop points out maybe internal connections to AMO (and sync?) may have
> filled up the limit of connections, not allowing to load more pages.
> If that's a possibility, I wonder how many connections do we open, and
> what's the timeout for those.

Thanks mak.

yeah - it sounds more like that than 692260. The similar bugs are 707744 and 407190.

I hope to work on the general concept of "better networking in crappy conditions" in the latter part of Q1, so this would be part of that. Stick [lame-network] in the whiteboard of anything else you think should be looked at in that effort.

Comment 6

[Marco Bonardo [:mak]]

If you have any script I may run in the Error Console when the problem happens, maybe it could help logging what's the problem.
Or I could just keep nspr http logging enabled till it happens...

Comment 7

[Dietrich Ayala (:dietrich)]

Are our internal connections (amo, sync, update, etc) persistent?

Comment 8

[Dave Townsend [:mossop]]

(In reply to Dietrich Ayala (:dietrich) from comment #7)
> Are our internal connections (amo, sync, update, etc) persistent?

For the extensions stuff I know we don't do anything to make them non-persistent (is there anything?) so I suspect they might be.

Comment 9

[Jan Tietjens]

Hi Guys,
this bug still exist in Firefox. Didn't checked the version, but must be one of the most current (24/25). 

My problem is, when I change the network from cabled to WIFI and vice versa the opened pages could not be loaded again. Opening new pages is also not working. I'm using Windows 7 64 Bit. With Windows XP there was no problem. 

This is really anoying as I often need to change the network. BTW. IE is working properly. Only a restart of the browser is fixing this error.

Could I do something to further analyze this error?

Thanks, JanT

Comment 10

[Jason Duell]

JanT: if you have concrete steps to reproduce this behavior (which URIs you opened) reliably, that would be great.

If it happens only sometimes, the next time it occurs take these steps, then attach the console output to this bug:

Go to about:config and find the network.http.diagnostics pref.. open
the js console (ctrl-shift-j) and clear it. Then flip the pref from
false to true (making it false to start with if necessary). That should
make a crude dump of http connection manager show up on the console.
This can be helpful in troubleshooting.

Comment 11

[Jan Tietjens]

Hi Jason,
thanks for the hint. I'll give you an update on Monday.

Comment 12

[Jan Tietjens]

Hi Jason,
I tried to receive the connection diagnostics, but nothing happend in the 'error' state. After restarting the browser I received the following properties:

[11:17:50.073] HTTP Connection Diagnostics
---------------------
IsSpdyEnabled() = 1
MaxSocketCount() = 542
mNumActiveConns = 0
mNumIdleConns = 0


Any ideas?

Best regards
Jan

Comment 13

[Jan Tietjens]

What can I do to further analyze this issue? Any ideas? Could I somehow reset the connection manager (if there is one)?

Comment 14

[Patrick McManus [:mcmanus]]

persona.jtiee I think you should follow 924432 and 939318 .. they are things we plan to work on in the next few months

Comment 15

[Jan Tietjens]

Ok, I'll follow them. Thank you Patrick.

Comment 16

[Patrick McManus [:mcmanus]]

the title of this bug is invalid according to comment 12.. I think its more likely we hit one of a number of resolved bugs over the last couple of years where the socket thread could spin or deadlock..

new information is best prusued in a fresh bug at this point - there isn't a lot of actionable value here.