Error: Socket Open Failed

[gumby_dammit223]

Hi,

Lately (ie. today) I've been getting a TON of these error messages:
[09:56:52] ERROR InterSocket - Error: Socket Open Failed, Host: news.giganews.com Error Code: 10060 A connection attempt failed because the connected party did not properly respond after a period of time, or established connection failed because connected host has failed to respond.

It started immediately after starting NB Pro 6.40 - Build 2059 and downloading headers. Checking the log for additional information I find a bunch of:
[10:17:50] HIGH - DG - Connect Flood Delay Active: GigaNews 7 secs Remaining before Reconnect

The timestamp doesn't mean there have been no errors or activity, just that I grabbed one from the almost 2000 in the list!

I even tried changing the host address to a straight IP address and still get:
[10:25:26] ERROR InterSocket - Error: Socket Open Failed, Host: 216.196.97.131 Error Code: 10060 A connection attempt failed because the connected party did not properly respond after a period of time, or established connection failed because connected host has failed to respond.

This error was preceded by:
[10:25:05] HIGH - GigaNews rec.pyrotechnics XOVER 247190-247200
|224 xover information follows [compress=gzip]

I believe I have been getting similar errors after upgrading to NB Pro from the GigaNews-specific version obtained quite a while ago, but I never really paid that much attention because they weren't so prevalent as they are today.

Any thoughts or suggestions?

Thanks in advance!
-gumby

[DThor]

First thought is that you're exceeding your maximum connections. Might want to log into the giga site and verify your account specs. Might want to change your password - perhaps your account has been compromised?

DT

[gumby_dammit223]

Wish it was that simple. I'm set up for 30 connections and that's what I'm using. I did change my password just to be safe, but it's still happening after changing it.

[Quade]

Dex has been noticing this more and more with the search server. He's not using Newsbin to feed the server. It's code that's probably 6 months old and unchanged. The frequency of these 10060 errors has skyrocketed lately to the point we had to change how we feed the search server. I don't really have an answer for you. A 10060 error is a networking error.

http://support.ipswitch.com/kb/WSK-19980714-EM32.htm

Meaning the connection times out. Newsbin is set to wait something like 60 seconds for the server to respond to a command and then the connection times out and Newsbin re-connects, If the connect fails immediately then you get the flood delay. Basically a failed connection waits 30 seconds before it tries again.

There's a setting you can add to the registry to increase the timeout value. Windows default is 4 minutes.

[itimpi]

Wish it was that simple. I'm set up for 30 connections and that's what I'm using. I did change my password just to be safe, but it's still happening after changing it.

You want Newsbin to be set up to use something like 5-10 connections LESS than your account allows or you will get this type of message. They occur because Newsbin for one reason can close a connection and then attempt to re-open it, and this reopen can happen faster than the server end gets around to cleanly releasing the connection and making it available for re-use. Having a small pool of unused connections gives a buffer for this scenario to be handled when the server is slow to recycle connections.

As a general rule you want to configure as few connections in Newsbin as you can and still use up your bandwidth on your link. Each connection has some protocol admin overheads so having more connections than you need can slow down over-all data throughput.

[gumby_dammit223]

Ok, thanks for the info, I'll check the link out shortly and try dropping the connection max down a bit. Strange though, that it's a problem that's only now breaking the surface.

Do you think it would do any good to bring this error up with Giganews support? Grasping at straws here!

I haven't thought to correlate my observations yet, but I _think_ I've noticed the error to pop up almost immediately if I download a single item, like a 100k picture. But, if I send it a relatively large download request, like a couple hundred part, multi-file download request, the error doesn't happen immediately. It's almost like the 30 connections are able to process a handful of downloads before things start backing up and choking. Again, this observation is far from scientific, proven or even confirmed, just a haphazard glace at a couple random times.

Thanks again!

[Quade]

If this happens occasionally but, the downloads continue, I wouldn't worry about it. If it happens enough to slow down your download, then it becomes a problem. It's not unusual to get disconnected from time to time. It might be worth reporting this to giganews. What's odd is I'm not seeing the same thing and I use giganews.

[gumby_dammit223]

A few dc's a session wouldn't bother me too much, but the fact I can't even download headers when I open the program and have to wait over a minute when it used to be almost instantaneous is ridiculous. After contacting Giganews, they suggested doing the network reboot thing and the disable virus software thing which did absolutely nothing. Even bypassing the router made absolutely no difference in eliminating the problem.

And regular internet browsing and ftp access doesn't seem to be affected in the slightest.

The frustration meter is edging into the yellow zone and am not sure what will bring it back into the green...except drinking heavily!

[DThor]

Just to reinforce itimpi's comment : keeping your connection count to less that the maximum is critical and I would argue 100% of the cause behind this. If you kept it at 25(less is better - only run as many as you need to max out your pipe otherwise you're 'network thrashing' on top of connect counts - actually slower! ) I'm suspecting this issue will go away. There's always Internet issues of one kind or another, an expectation of no disconnects probably isn't realistic. I really wish providers would make this clear, especially the ones that market by throwing large connection numbers at you. But then, I guess that wouldn't be marketing.

DT

[Quade]

Well, he's reporting an initial issue and then it clears up. I'm going to look into that. I can say I don't see it here with 2049. I might have seen it with earlier version.

[gumby_dammit223]

I throttled back the connection number from 30 to 20 and no difference. I rebooted every piece of equipment between me and the internet (computer, router, cable modem) and still no difference. I ran AVG PC Tuneup and let it optimize my network settings and basically clean up the computer and still no apparent change in error generation.

I have NB Pro v6.21 build 1534 running on another computer, a much older and slower one and I believe I am seeing similar errors.

Per a suggestion from Giganews, I tried changing from SSL port 563 to SSL port 443 and again, no difference seen. I also tried using the newsreader in Outlook Express and I think I saw similar problems, but without any tracking, I don't know what was going on, except to get a handful of errors about not connecting to the server.

The one thing I did notice today was interesting. There are 20 newsgroups that I monitor. When I loaded up NB, all 20 showed up in the download list for a header refresh. The first 10 seemed to download like I would expect, that is within a couple seconds. The next 10 sat there for a minute or so, and then the error messages popped up. Then, over the next couple of minutes as errors timed out and retrys began happening, the header refreshes slowly completed, in a onesy twosy type of process. Never a mass update like I had been expecting.

Because it's happening on two radically different computers (both running WinXP but with a 10 year age difference) I am pointing to something possibly ISP-related, but I have nothing definitive I can accuse them with.

Any other ideas?

[gumby_dammit223]

Ok.....I just put the system through a little stress test by downloading a 100 MB file. I didn't see any hiccups through the vast majority of the download. My d/l speed climbed to >16 Mbps and stayed there which is what I expect to see. I had all 30 connections enabled.

I'm beginning to think I might need to focus my ire on my cable company for doing something to foul-up my network. Any ideas how I can prove/disprove this???

[driverdude]

It is the new stuff the ISPs are installing to smart inspect connections for UP LOADER violations of copyrighted works for the 6 strikes initiative this year in the usa.

That is my 85% hunch. I could be wrong, it could be aftermath of errors from the MASSIVE amazon datacenter outages today in afternoon affecting much of internet but happened before then, and after.

His problem started with me the same day, but even after unplugging cable modem, changing my MAC address (to get a new IP), waited 20 minutes, then got a new IP, a few times in the day later on with lots of activity I got those.

IT IS ON GIGANEWS ***AND*** ASTRAWEB about equal per i/o as percentage , it is NOT nntp supplier fault it is happening between me and the the usenet suppliers I think.

for ASTRAWEB I was using 8 connections and NOT using SSL

for giganews i was using a pathetic 15 connections and standard SSL

issues for timeouts happen sporadically now and then day or night since the ISPS enabled the "6 strikes and you are out" copyright infringement detector for UPLOADERS a few days ago

I was using a tool called "netbalancer 5.2.1" (a no timebomb no-ad freeware throttler that is very very precise) to set my I/O to 74% of my contracted business line data rate. My ISP (comC$$t) uses 75% in an algorithm publicly stated twice officially, as the steady abuse threshold. The line can go over triple that speed if not throttled but result in calls from the IT section to me after only one day or so of "abuse" even with over a week of 100% NONACTIVITY before the ballsout test. I used to set to a lower number in newsbin but newsbins throttler is designed for very high datarates i think, and the scheduler interferes with the pause button when i try to pause BEFORE i hit quit for safety when i try to quit.

Anyway... I got EXACTLY the same issues starting the same day, but 6.4 was fine for weeks under the same load prior to that day.

Years ago, when I had dropped connections that got retried, I had a ping route time that seemed to go from USA to india and back... and the next morning I got a call from ISP it callcenter in india for "abuse" I also had for the very first time, an issued IP that was a unique C block I never had seen

Here is a Oct 2012 "fluff" news article link on the 6 strikes ISP initiative and the new code on the isps backend, that apparently is an overworked box that at times I guess drops NNTP and SSL connections :

http://www.ibtimes.com/how-six-strike-p ... bat-online

The only thing i use usenet for is reviewing pre 1925 content so no concerns on my end

========

example messages (a dozen per day since the beginning of this mess a few days ago)

"'ERROR NNTPSocket - Server:xxxxxxxxxxxx NNTPServer: Timeout - Waiting for Server to send Data: Socket - Read or Write Timeout"
and
"ERROR Socket - Error: WSARecv Socket Error Code: 10060 A connection attempt failed because the connected party did not properly respond after a period of time, or established connection failed because connected host has failed to respond."

and

"InterSocket - Error: Socket Open Failed, Host: news.astraweb.com Error Code: 10060 A connection attempt failed because the connected party did not properly respond after a period of time, or established connection failed because connected host has failed to respond."

I could be mistaken but I do not think it affected my downloads overall, I should have done more research, but I do have a debugging log from one day.

my machine was running nothing at all else because that box is essentially dedicated to the amazing program entitled Newsbin Pro 6.40

[Quade]

I'm skeptical this is related to the 6 strikes thing. From your own article, they're just going after P2P. Usenet isn't P2P. I just think it's regular connectivity issues. The fact it happens at random times also suggests it's not some attack. It's happening at our data center too suggesting it has nothing to do with the ISP too.

[driverdude]

I'm skeptical this is related to the 6 strikes thing. From your own article, they're just going after P2P. Usenet isn't P2P. I just think it's regular connectivity issues. The fact it happens at random times also suggests it's not some attack. It's happening at our data center too suggesting it has nothing to do with the ISP too.

You totally misunderstood my point.

All boxes that analyze packets can have adverse effects no matter what the desired goal.

I never claimed they tried to do anything that would be detected, or adverse, or legally endanger them again to a SECOND 16 million dollar federal court judgement against them.

I am merely insinuating that the reason both astraweb and giganews both had a FEW dropped connections equally here on my end probably has to do with the "six strikes" MarkMonitor technology that is in the MIDDLE of all this traffic that just went live for me here, not keeping up.

I am stating that the errors are accidental side effect of equipment not ready for prime time. In fact, for all I know its not even being used fully yet precisely because of issue of packet sniffer not keeping up.

I assume this six strike MarkMonitor technology thing if installed in partnered ISPs needs to interfere with all traffic, not just what they claim is P2P, and in fact probably does have potential to interfere with a lot as a side effect.

Officially it goes 100% full live for all ISPs and every end user Nov 28th 2012.... but they obviously need to install and test it and are possibly the cause of these issues

Nov 28th rollout mention :
http://www.p2pon.com/2012/10/19/att-to- ... es-system/

MarkMonitor Testing to commence around October 20th 2012 :

http://www.dailytech.com/Piracy+Warning ... e27996.htm

six-strikes MarkMonitor in a press release from their OWN web site bragging that it starts testing after October 18th 2012 for certain
[warning do not click this link normally:]
http://www.copyrightinformation.org/node/709

If you see these connection issues using backbones and colocenters to giganews and straweb, then it is obviously some OTHER strange disruption, or a coincidental unrelated issue and markmonitor six strike early testing of boxes issue still remains plausible.

[gumby_dammit223]

I've been trying a lot of different things per Giganews and one of the last things they suggested was trying their European server. After several attempts connecting to their US server and experiencing the same errors, I connected to their European server and had ABSOLUTELY NO CONNECTION ERRORS whatsoever. I let Giganews know this a few minutes ago so we'll see what they come back with. In the meantime, I guess all my newsgroup activity will be from Europe!

[Quade]

We're seeing issues with the US server too.