ICQ - connection reset by peer

[babaz]

Hi, i have a problem, pidgin can't connect to icq, and return this message: "Lost connection with server: Connection reset by peer". Msn, gmail account works perfect. Thanks and sorry for bad english.

[JuliuX-Arg]

Replying to babaz:

I have the same problem, only with ICQ. Additionally it hangs my modem!!! (I have to turn it off, and on again).

[MarkDoliner]

Can you paste the output from the debug window when this happens?

[babaz]

the problem is resolved.

[tinloaf]

I'm reopening this ticket because this problem happened to me a few times lately - I'm attaching the debug output.

[MarkDoliner]

tinloaf: You say it's happened a few times lately... is it still happening now? Are you able to connect, but then you get disconnected after a few minutes? Have you ever been able to connect? Are you connected through a wireless network?

[tinloaf]

Mark: Yes, it happened to me right now.

I'm able to connect, the buddylist starts being built, but right after one second or so I get the "connection reset by peer". I have to try several times (usually that many that i get the "Wait 10 minutes until next login"-error once) until I can login.

Yes, I have been able to connect for several years, and I am still able to connect now, it just costs me half an hour or so.

Yes, I am connecting through a wireless network, but this is stable, i checked that. Also my other accounts (MSN, Jabber) are no problem to login.

Judgeing from the log it's the following problem (but i don't really now):

When Pidgin logs into ICQ it tries to update every single buddy in the list and does that in a very fast way - that fast that the server disconnects us. There are "rate warnings" from the oscar module in lines 126-130,770, and 818-831 in the attached log, the last ones being directly before the "The last action could not be executed because the rate limit was exeeded"-messages in lines 832-836 and 858-876.

[tinloaf]

PS.: Look into oscar's oscar.c file, function "purple_parse_ratechange(...)" - this function is called if a "rate limit" (or any other send-rate related) packet is received on FLAP and displays the messages above, but does not really apply a limit. I don't know how to implement a rate limit in libpurple without completely freezing it, otherwise i would try to...

[MarkDoliner]

tinloaf: Can you check Pidgin's About window and verify that you are using Pidgin 2.3.1? Recent versions of Pidgin already rate limit themselves, so in practice you should never exceed the rate limit.

What do you mean by "tries to update every single buddy in the list"? What would Pidgin be updating, exactly?

Also can you check if you're running any 3rd party plugins? You can see which plugins are loaded by going to Tools->Plugins in the buddy list and selecting "Plugins." Thanks

[Bock]

The same trouble at 2.4.0 Log is attached. I started pidgin, it connected to server, list is building and "connection reset by peer". It was at 14 march and the same oing now.

[Bock]

log of "connection reset by peer"

[Bock]

It's causes if too many peoples from contactlist in online. Now, at morning, I connect without troubles, but I think that at 15-00 I get "connection reset by peer".

[mweinelt]

I actually have the same issue with 2.4.1, just built. My ICQ contact list is about 500 people long. Any help i can get you on this issue?

[Bock]

I think that pidgin don't make anythink then get "oscar: rate limit" and I think that pidgin must limit speed of update. I have about 350 people in contact list.

[mweinelt]

reopen perhaps? cmon ppl, i can't use icq for like a week now, can't stand this any longer.

[mweinelt]

uhm? this problem is quite serious and _not_ minor in my opinion. if only a few ppl are online i can connect to icq without problems, but in the rush-hour where like 150-200 people are online i'm getting kicked out directly after the contact list is display with the message "Connection reset by peer".

[deryni]

Ignore the Priority value it doesn't mean what you think it does.

[Bock]

I think too, that if I can't make normal connect at day hours and chatting with people - it's not minor..

[deryni]

I made no comments about the severity of this bug, or how much it affects you. My comment was that the Priority field on this ticket doesn't mean what it was implied to mean.

[mweinelt]

Tracked the issue back to the login-server. Changed Host/Port? to login.icq.com:443 instead of the aol host.

[Bock]

It's not helped. Log is needed?

[maniax]

Debugging this, it seems like the problem is that the rate changes send from the server are not honored in the initial buddy list initialization.

aim_icq_getstatusnote() in libpurple/protocols/oscar/family_icq.c gets called too much, and seems like it's creating the problem. I tried an ugly hack by adding an usleep() in it to sleep for about 300ms, and it connected on the second try, but that seems to be too ugly even to think about leaving.

Maybe the problem is that aim_icq_getstatusnote() uses flap_connection_send instead of flap_connection_send_snac() (which does queueing) ... I'll have to dig a bit deeper.

I can provide packet captures, it anyone is interested.

[maniax]

Another note - I don't have the issue with 2.3.x, but it started showing at 2.4. I have 360 contacts in my ICQ, and about 55 are online (or that's the number at which my connection dies, according to the debug log).

[maniax]

Ugly hack to show a working (but not production-ready) solution to the issue, use on your own risk.

[MarkDoliner]

Oh, Evan Schoenberg fixed this within the last week or two. The solution was to change the oscar code to always send SNACs using the appropriate function for rate limiting, instead of sending SNACs directly. The fix will be in Pidgin 2.4.2, when it's released. Until that time you might want to just downgrade to 2.3.x.

[jaaa]

i have _STILL_ problems with Pidgin 2.5.0 (and i had this problem also with 2.4.2). i have two accounts and i can connect only with one - the second gives Lost connection with server: Connection reset by peer . but it's possible to connect with Pidgin 2.3. i'm attaching debug log