open All Channels
seplocked Issues, Workarounds & Localization
blankseplocked Investigating Socket Closed Errors - How You Can Help
 
This thread is older than 90 days and has been locked due to inactivity.


 
Pages: first : previous : ... 8 9 10 11 12 13 [14] 15 16 : last (16)

Author Topic

Bugszor
Caldari
PodPal
Leathermen
Posted - 2010.08.31 23:46:00 - [391]
 

This should rather be fixed fast..

i just paid for another month after noticing for soe reason i missed my payment >_<
and i cant even frigging log in!

it either.. hangs in authenticating. till it says sockets closed..
or hangs at char select screen..

OR it hangs at my ship without any chat or overview. and ofc again socket closed..

How can this suddenly out of nowhere hit anyone?
and this issue has been here for 2 years? and still no fix?
come on CCP. this is some bad effing juju here..
Yer gonna loose people.. mark my words...

Anddeh McNab
Cadre Assault Force
Posted - 2010.09.01 09:08:00 - [392]
 

Recently been suffering from these errors, in the last day or so. At least I think they're socket closed errors, the client just seems to disconnect and I'll be unable to do anything. Which usually normally means me CRTL+Q'ing in fury because the client seems to take forever and a day to actually realise it's been disconnected from the server.

Steve Celeste
Overdogs
Posted - 2010.09.02 20:10:00 - [393]
 

Originally by: Scouteye
Remove uTorret and its buddy app DNA.

This fixed it for me, thank you.

I wasn't getting the classic "socket closed" errors, but rather the one where you are disconnected and your client doesn't even realize it.

Steve Celeste
Overdogs
Posted - 2010.09.04 09:13:00 - [394]
 

...and it's back.

**** this game.

Shuckstar
Gallente
Hauling hogs
Swine Aviation Labs
Posted - 2010.09.14 01:12:00 - [395]
 

Started getting this last 2 days, never had it before, as for loggin on sisi and recreating it, why dont you want the logs and bug report from tran not the test server?

Tbh its a joke it happens at random times, undocking, in middle of mission, pvping whatever im doing it just comes up socket closed. Been a loyal customer since 2005 and it not my job to log in on test server and waste more time on there, especially when can give you logs from tran client :S

At moment game is unplayableTwisted Evil, ill bug report it from main server not tranq. If carry's on like this and according to ccp fallout this has been going on since November 2008? Not to suprised to be honest as it takes you years to fix anything. Anyway that's my rant over. If keeps happening i wont be renewing my accounts again, and nope no one getting my stuff.


MasterOrg
Posted - 2010.09.14 11:43:00 - [396]
 

I have been experiencing these also. They had stopped but they started back up. I get dropped out of all 3 windows sometimes, and sometimes its just 2 of 3 windows. It is quite annoying. If you aren't going to address the issue, at least edit portions of the game, so that those of us affected, aren't taken advantage of.

One night I would connect, and then I would get disconnected, I'd reconnect and thirty seconds later I was disconnected. It got so bad that I just shut down EVE and went to bed.


HomeGrownBudZ
modro
R.A.G.E
Posted - 2010.09.17 07:30:00 - [397]
 

Socket closed just cost me 250Mill.....and i just renewed 1 year TODAY....thank you CCP.... and yes, much QQ

Ghennings
Posted - 2010.09.17 07:52:00 - [398]
 


I had the lag and socked closed error thing for sometime. I updated the firmware from my adsl modem and that did it for me. I hope this little info helps someone else.

Amarant'h
Posted - 2010.09.18 18:30:00 - [399]
 

Been having this bug like 2 weeks or so. ANNOYING!!! Its just random 'socket close' error message and game quits. This happens many times in one hour. Cant do anything. No fix anywhere and ive tried everything.

horza tuu
Posted - 2010.09.19 21:10:00 - [400]
 

Glad to know this is not just me. Hope this gets sorted out soon. My main char is a trader, so if I'm busting through a few jumps with some pricey cargo, and I get disconnected at a gate, then I'm in gank city.

Calandari
Retribution. Inc.
Strategic Operations Brigade
Posted - 2010.09.24 07:10:00 - [401]
 

this is ******ed .... I keep getting socket closeures ... really come on ... ever 10minutes

chacoty
Posted - 2010.09.26 19:09:00 - [402]
 

bin having this trouble for a few weeks now disconnects every 30 minutes or so really ****ing annoying wen in fleet hope it gets sorted soon

Marketing Agent
Posted - 2010.09.30 09:24:00 - [403]
 

I just fixed my IP adress to the router, as in before I have always let the computer assing it itself, now I have a static one, and now pingplotter have no problems what so ever with 40% packetlosses at the last server on the list (CCPs server). I have yet for 10h to make a single random disconnect in the game. Makes no sense but welp, there it is.

Jaari Val'Dara
Caldari
Deep Space Nomads Corp
Posted - 2010.10.01 16:22:00 - [404]
 

It seems EVE client doesn't like packet loss very much. Sometimes I can play just fine, but sometimes it keeps disconnecting, while everything else works fine. Hell, I talk on TS without a hiccup, while EVE disconnects. EVE shouldn't declare me dead if one packet goes the other way.

Wojomon
Posted - 2010.10.02 03:08:00 - [405]
 

I have only played for a couple days but liked the trial so decided to pay for a year. After the patch I now can't stay on for more than 5 mins at a time before getting this error.

Holly Hoover
Posted - 2010.10.12 00:22:00 - [406]
 

It's not packet loss, or anything else. It's their servers and they need to fix them!

Cutter Isaacson
Minmatar
Spycotics
Posted - 2010.10.14 02:17:00 - [407]
 

I've been doing a bit of noobish research into socket closed errors on various systems and pieces of software and one thing keeps cropping up, that socket closing errors are usually a result of bad coding at the database end, rather than the client end.

One thing in particular keeps cropping up regarding packetloss and subsequent disconnects and that is that if there is an interuption in the flow of data, the database server interprets that loss, however brief, as the client disconnecting. There doesnt appear to be any sensitivity setting on CCP's database server, or rather in the microsoft sql software they use, which given its complexity in other areas is something of a suprise.

The only other major cause of socket closed errors i could find regarding the sql software is a memory overload at the socket itself, where the queue of data trying to access that socket becomes too much for it to handle and it closes, resulting in disconnects for us. Now taking into account the number of people getting disconnects and the various OS and hardware settings in use by the players, it seems highly unlikely that the problem is at our end.

Thus the only conclusion i can draw is that it isnt us, or our hardaware or individual ISP's but rather a hardware or coding error at CCP's end. Can we please get some feedback on this from a Dev as soon as possible.

Borun Tal
Minmatar
Space Pods Inc
Posted - 2010.10.14 03:02:00 - [408]
 

My last post here appears to have vanished, so I'll post again:

Prior to last expansion, I went years with maybe one d/c per year. Since last expansion, I've been getting several d/c's A NIGHT. Got bad after "optional" patch #1, and tonite I've d/c'd twice and been on for only an hour or so. Local network activity is virtually non-existant, internet packet traces while client running is very acceptable, and other network apps (such as RDC to work computers and Positive Networks connections) working witout a hitch. Load is not an issue.

And the Eve client craps out, sometimes a Socket error, sometimes just a crash without an error.

CCP: Paying clients here. Plan to at least comment beyond the "we're working on it"?

Holly Hoover
Posted - 2010.10.14 04:47:00 - [409]
 

Originally by: Cutter Isaacson
I've been doing a bit of noobish research into socket closed errors on various systems and pieces of software and one thing keeps cropping up, that socket closing errors are usually a result of bad coding at the database end, rather than the client end.

One thing in particular keeps cropping up regarding packetloss and subsequent disconnects and that is that if there is an interuption in the flow of data, the database server interprets that loss, however brief, as the client disconnecting. There doesnt appear to be any sensitivity setting on CCP's database server, or rather in the microsoft sql software they use, which given its complexity in other areas is something of a suprise.

The only other major cause of socket closed errors i could find regarding the sql software is a memory overload at the socket itself, where the queue of data trying to access that socket becomes too much for it to handle and it closes, resulting in disconnects for us. Now taking into account the number of people getting disconnects and the various OS and hardware settings in use by the players, it seems highly unlikely that the problem is at our end.

Thus the only conclusion i can draw is that it isnt us, or our hardaware or individual ISP's but rather a hardware or coding error at CCP's end. Can we please get some feedback on this from a Dev as soon as possible.


That is the most plausible explaination I've read.
I also suspect that there is no solution because it is inherent in the current setup which reduces lag by dropping clients!
BTW the forum coding is bugged.

Borun Tal
Minmatar
Space Pods Inc
Posted - 2010.10.14 18:34:00 - [410]
 

Originally by: Cutter Isaacson
Thus the only conclusion i can draw is that it isnt us, or our hardaware or individual ISP's but rather a hardware or coding error at CCP's end. Can we please get some feedback on this from a Dev as soon as possible.


Good luck with that. Smile

Cutter Isaacson
Minmatar
Spycotics
Posted - 2010.10.15 14:05:00 - [411]
 

/me pokes CCP for a response. Come on guys and girls, you ask us to test your game for you, give us a test server to go on, how about a few updates here and there? After all this thread is now 14 pages long and counting, some responses to querys raised here would be greatfully received Very Happy

CCP Warlock

Posted - 2010.10.22 14:06:00 - [412]
 

Before I get into the nitty gritty on this, could I ask that if you are having persistent socket closed problems, please raise a petition. That will let the GM's help you with some of the more frequent causes, and also help you get us the information we need to try and debug these problems. Many thanks to all of you who have done that, and especially the people who have sent in wireshark traces, because it really helps us try and sort out these problems.

In terms of things that can't cause this. Cutter Isaacson is correct that Database connectivity can be an issue for some systems. However, the Eve client doesn't connect directly to the database, it connects directly to a proxy server, which then routes requests around the cluster, so that isn't a source of connectivity issues for us.

We do not deliberately drop clients when we have lag. We sometimes drop them accidentally because of software bugs, and you can also get dropped if the Sol or the Proxy that you're currently using crash. None of these things should cause repeated disconnections though, they're just one offs, and node failure is fairly rare in the Cluster these days.

If you suddenly start experiencing this issue, and especially if it's a weekend, it's quite likely that it's a network provider somewhere along the connection. Most network maintenance is done at weekends. Similarly if you get a disconnect at the same time every day, or even hour. Some ISP's reissue their IP addresses every 24 hours, and that will also drop the connection.

Check your file sharing software, and make sure it's throttled. I think about half the wireshark traces I see, file sharing congestion is at least implicated. File sharing can be an issue because it uses UDP. Eve uses TCP - the reason this can cause problems is that UDP is an unreliable protocol and TCP is a reliable protocol.

Now this is a fun one. In order to be a reliable protocol (bear in mind this is the Computer Science definition of reliability which translates to all your packets will be delivered reliably and in order, until they can't be when the connection will be dropped), TCP has some adaptive features that allow it to back off if it thinks the network connection is being blocked by too much traffic. UDP however doesn't care, and will just hammer packets through continuously. The result can be that with a lot of UDP traffic, TCP traffic will get starved of network bandwidth, and eventually the connection will be dropped. The EVE client has a very low packet rate, which doesn't help with this problem.

This is also why idling was a problem. This should be fixed now, but we identified a problem last year where some home routers were disconnecting players, because they hadn't seen any traffic on the Eve link for some time. We've turned the TCP KEEP_ALIVE socket feature on on our side to address this.

Several people have commented, and I'm inclined to agree after looking at a bunch of wireshark traces, that the EVE client is too sensitive to occasional hiccups on the link. We're looking at that at the moment, and seeing if we can change some of the TCP settings on our side to be a little more tolerant.

Tldr; Raise a petition. There is a lot of things that can cause this issue, both on your side and on ours, and if you're having it repeatedly we'll need to work with you to go down the list, and sort it out.



Bremen Nacht
Minmatar
Posted - 2010.10.22 16:04:00 - [413]
 

Originally by: CCP Warlock
Before I get into the nitty gritty on this, could I ask that if you are having persistent socket closed problems, please raise a petition. That will let the GM's help you with some of the more frequent causes, and also help you get us the information we need to try and debug these problems. Many thanks to all of you who have done that, and especially the people who have sent in wireshark traces, because it really helps us try and sort out these problems.

In terms of things that can't cause this. Cutter Isaacson is correct that Database connectivity can be an issue for some systems. However, the Eve client doesn't connect directly to the database, it connects directly to a proxy server, which then routes requests around the cluster, so that isn't a source of connectivity issues for us.

We do not deliberately drop clients when we have lag. We sometimes drop them accidentally because of software bugs, and you can also get dropped if the Sol or the Proxy that you're currently using crash. None of these things should cause repeated disconnections though, they're just one offs, and node failure is fairly rare in the Cluster these days.

If you suddenly start experiencing this issue, and especially if it's a weekend, it's quite likely that it's a network provider somewhere along the connection. Most network maintenance is done at weekends. Similarly if you get a disconnect at the same time every day, or even hour. Some ISP's reissue their IP addresses every 24 hours, and that will also drop the connection.

Check your file sharing software, and make sure it's throttled. I think about half the wireshark traces I see, file sharing congestion is at least implicated. File sharing can be an issue because it uses UDP. Eve uses TCP - the reason this can cause problems is that UDP is an unreliable protocol and TCP is a reliable protocol.

Now this is a fun one. In order to be a reliable protocol (bear in mind this is the Computer Science definition of reliability which translates to all your packets will be delivered reliably and in order, until they can't be when the connection will be dropped), TCP has some adaptive features that allow it to back off if it thinks the network connection is being blocked by too much traffic. UDP however doesn't care, and will just hammer packets through continuously. The result can be that with a lot of UDP traffic, TCP traffic will get starved of network bandwidth, and eventually the connection will be dropped. The EVE client has a very low packet rate, which doesn't help with this problem.

This is also why idling was a problem. This should be fixed now, but we identified a problem last year where some home routers were disconnecting players, because they hadn't seen any traffic on the Eve link for some time. We've turned the TCP KEEP_ALIVE socket feature on on our side to address this.

Several people have commented, and I'm inclined to agree after looking at a bunch of wireshark traces, that the EVE client is too sensitive to occasional hiccups on the link. We're looking at that at the moment, and seeing if we can change some of the TCP settings on our side to be a little more tolerant.

Tldr; Raise a petition. There is a lot of things that can cause this issue, both on your side and on ours, and if you're having it repeatedly we'll need to work with you to go down the list, and sort it out.





Thanks for the update, glad to know you guys are looking into this

Cutter Isaacson
Minmatar
Spycotics
Posted - 2010.10.24 00:52:00 - [414]
 

Originally by: CCP Warlock
Nicely detailed response Very Happy


Thankyou Warlock, for your reply. It's nice to know the things I got wrong so from now on I won't waste my time looking in to those areas. I intend on looking more into issues that might actually be being caused at my end now that I know what I am looking for.

As for the TCP/UDP issue, is there anything that can be done about this? Either by CCP or by me? Or is this an issue with all ISP's and network infrastructures as a whole? Also on the issue of EVE being sensitive to hiccups, is there any way to increase my connection "strength" to the proxy server much in the same way a signal booster helps with analogue TV reception?

Once again thankyou for the update, its nice to know threads like this are followed Very Happy


Keep up the good work and we look forward to being kept informed on this subject.

Cutter Isaacson
Minmatar
Spycotics
Posted - 2010.10.24 00:58:00 - [415]
 

Edited by: Cutter Isaacson on 24/10/2010 01:00:24
Actually Warlock, I have a few other questions. You say that the client connects to a proxy and not directly to the database, which in hindsight I should have thought about. What I would like to know is:


1. How exactly does the client connect to the proxy?
2. What protocols does the proxy have for dealing with each incoming data packet from the client software?
3. Could there be a similar issue with the proxy as the one I described for databases?

Please excuse me if my questions seems rather naive but I am new to all this and I am learning as I go.

Once again many thanks.

Edited for poor keyboard control resulting in spelling errors Sad

CCP Warlock

Posted - 2010.10.25 21:22:00 - [416]
 

First on the TCP/UDP issue.

We do discuss this from time to time, since essentially either protocol has a list of advantages/disadvantages which are typically orthogonal to each other. Using UDP might make some momentary glitches easier to disguise, but we could achieve a similar goal just by putting in some queuing semantics and trying to reconnect over TCP, whilst quietly not reporting the problem to the Eve Client. There is a fear that this might open up some griefing/hacking opportunities though, which has made us hold back on that one. One distinct argument against using UDP, is that because UDP is not network sensitive, overloaded internet routers will preferentially drop UDP packets, in order to prevent UDP traffic starving out TCP. This would potentially make any congestion issues even worse.

The client connects to the Proxy over a normal TCP/IP connection. We keep that up for the whole session. Now TCP is a stream protocol, which means that it just sends/receives data continuously. In practice, most applications including Eve will break up their data into messages that are sent in send/response pairs. What happens is that between the proxy and the client, and the proxy and any machine in the cluster, we provide a message layer, which the machines use to communicate with each other. From that perspective TCP is just a socket address that we send a message to, and can expect responses from.

This has the advantage that we can protect players from a fair amount of database lag, simply by handling that at the application layer.

So yes, although it's possible that there could be a similar problem to the one you described with databases on the proxies we aren't seeing any evidence of it. We monitor them pretty carefully, and always run them at a much lower packet rate than they're capable of handling. CCP Veritas also beefed up the TCP statistics reporting on them this summer, so we're as confident as we can be that they're running clean.

In terms of troubleshooting - the trouble with network issues is that literally anything can be a source of problems, at both ends. Wireshark traces from the player are invaluable for finding the proximate cause of the disconnection, but often what it shows is a series of TCP retransmit requests, no response from the server, and then TCP tears down the connection, as it is supposed to. We don't know from that where or why the connection was dropped.

Now if we see a lot of other errors around the same time, or a lot of UDP traffic, or just periodic bursts of errors, then an educated guess will be it's likely to be something closer to the player than us. It's a lot easier in practice to ignore the occasional web page not loading, or getting frozen than the Eve client putting up a socket closed message, even though the Web page not loading is often also some kind of intermittent TCP problem.

However, as I said, my network troubleshooting spider sense is really not happy about this whole issue, and we are continuing to investigate this on our side.


Bremen Nacht
Minmatar
Posted - 2010.10.26 20:07:00 - [417]
 

Originally by: CCP Warlock
First on the TCP/UDP issue.



However, as I said, my network troubleshooting spider sense is really not happy about this whole issue, and we are continuing to investigate this on our side.




If this is correct, why has my petition just been closed?

Pheriann
Posted - 2010.10.27 02:50:00 - [418]
 

I have been having this issue for the last year or so. I have petitioned 3 times now and not gotten anywhere. On my latest petition I was told that this is a connection issue. So I decided to look into it myself a little.

Quote from my logserver file:

494010 2010.10.27 01:06:45:927 Creating Closed Packet: The socket was closed exception:error(10054, 'Connection reset by peer')

494011 2010.10.27 01:06:45:928 Couldn't send close packet, rhe socket is probably already closed. [Errno 10054] Connection reset by peer: 'WSASend failed with error 10054: An existing connection was forcibly closed by the remote host.\r\n'

494012 2010.10.27 01:06:45:947 Received OnDisconnect with reason= 0 and msg= The socket was closed

494013 2010.10.27 01:06:45:948 GameUI::OnDisconnect 0 The socket was closed

494014 2010.10.27 01:06:45:948 About to reboot the client

It is implied that it is my internet connection at fault. Now a WSASend error is a DataBase error. As per Microsoft :

Connection reset by peer.

An existing connection was forcibly closed by the remote host. This normally results if the peer application on the remote host is suddenly stopped, the host is rebooted, the host or remote network interface is disabled, or the remote host uses a hard close (see setsockopt for more information on the SO_LINGER option on the remote socket). This error may also result if a connection was broken due to keep-alive activity detecting a failure while one or more operations are in progress. Operations that were in progress fail with WSAENETRESET. Subsequent operations fail with WSAECONNRESET.

1: Application on Host is suddenly stopped = False! (Well at least not by me.)
2: Host is rebooted = False!
3: Host or remote network interface is disabled = False!
4: Host uses a hard close = False!
5: See setsocketopt for more info = Can't it is not on my side!
6: Keep-Alive detected a failure = Can't check again because it is not on my side!
7: Progress fail with WSAENETRESET = Can't be this cause this is error code 10052!
8: Subsequent operations fail with WSAECONNREST = Well dah this is error code 10054!

So what I am trying to understand here then is this:

1: EVE's database interaction says that the sockets to my system are closed.
2: My router and modem are up and active.
3: I can access the internet.
4: I can d/l files.
5: I can stream video.
6: I can listen to internet music.
7: I can play other games.

So how is that possible? If I can do 2-7 how is 1 correct? If I can do 2-7 all at the same time and have no issues whats up with 1?
I find it very hard to believe that it is on my end if the only application that has this issue is EVE.

One last thing, and I maybe out in left field on this but, if this is a winsock error why is ccp looking at the tcp/ip layer?

Just my rants/flame/questioning....2 cents!





Tarin Majagh
Caldari
Caldari Provisions
Posted - 2010.10.28 06:06:00 - [419]
 

What is the likely turnaround on petitions? I have an open one, and for the last few days its been left open with no response. I don't need to remind you that we're paying customers, and currently you're not providing a service we're paying for...

Its been over a month since I've been able to play this game, and its starting to get a bit of a joke now. Surely fixing the already existing problems would be a better idea instead of introducing new features and functionality...

Holly Hoover
Posted - 2010.10.28 13:25:00 - [420]
 

I'm glad there has been a detailed response so thank you.
As a layperson though, most of it is above me.

However it seems to me my issue seems related to lag or load. The odds of getting disconnected are very high in Jita when docking and undocking at the heavily camped hub station. A heavily camped gate also creates frequent disconnects. To me that seems like a server issue no?


Pages: first : previous : ... 8 9 10 11 12 13 [14] 15 16 : last (16)

This thread is older than 90 days and has been locked due to inactivity.


 


The new forums are live

Please adjust your bookmarks to https://forums.eveonline.com

These forums are archived and read-only