Connection drops when connecting to remote server

Windows 8.1 ITA 64bit

When I connect to a specific remote server (hosted on Amazon) my connection drops after about 60 seconds. That is:
- Putty: after 60 seconds (idle / non idle) connection drops (PuTTy window status: inactive, PuTTy popup error: "Network error: Software caused connection abort")
- RockMongo: if I connect using RockMongo, after 60 seconds connection is lost
- Jmeter: about every minute I have "Non HTTP response code: java.net.SocketException,Non HTTP response message: Socket closed"

I have not this problem when connecting to other servers BUT my friends, connecting from the same office using their Linux OS, don't have this problem.

I have the same problem with wired (Atheros AR81xx) and wireless (Atheros AR956x) connection. Wired and Wi-Fi drivers are up-to-date.

No antivirus or firewall installed. I tried turning windows firewall on and off, nothing change. I also tried to set MTU to the best value for my connection (1472) but this didn't change the game. I can't say if the problem existed in the past, I noticed it only few days ago when I started to work with that server.

I also tried to dump tcp packets using netsh, I got some errors (Winsock - event id 1032 ad 1003) and I got these errors also closing all opened programs.

Any help?

Have you talked to Amazon S3 group, or looked in the discussion forum at https://forums.aws.amazon.com/forum.jspa?forumID=24 to see if others may also be having the same problem, or had it and found a solution.

Instead of using PuTTY, try WinSCP. It incorporates Putty, a SSH & FTP Client all in one. You can download it from WinSCP :: Free SFTP and FTP client for Windows The only thing that I can think of, is that the S3 server is dropping your connection on their end. WinSCP also allows for logging, so that you are able to better troubleshoot what is going on. More on the logging feature. Logging :: WinSCP

I'll talk to Amazon s3 group too, thank you.

This is WinSCP log:

. 2014-09-19 10:31:53.696 Startup conversation with host finished.
. 2014-09-19 10:32:42.007 Changing directory to "mydir".
> 2014-09-19 10:32:42.083 cd "mydir" ; echo "WinSCP: this is end-of-file:$?"
. 2014-09-19 10:32:57.084 Waiting for data timed out, asking user what to do.
. 2014-09-19 10:32:57.250 Asking user:
. 2014-09-19 10:32:57.250 **Host is not communicating for 15 seconds.
. 2014-09-19 10:32:57.250
. 2014-09-19 10:32:57.250 Wait for another 15 seconds?** ()
. 2014-09-19 10:33:14.253 Data has arrived, closing query to user.
. 2014-09-19 10:33:14.527 Network error: Software caused connection abort
* 2014-09-19 10:33:14.586 (ESshFatal) Network error: Software caused connection abort
* 2014-09-19 10:33:14.586 Error changing directory to 'mydir'.


I have the same problems also with jmeter and rockmongo and they use http connection. Consider I have no problem using Linux from the same office with same url & account, so it "seems" a OS related problem.

Other ideas?

The problem exists only versus Amazon VPC; no problem using Amazon EC2. Maybe it's a routing problem. Our company use a VPN to access to Amazon VPC, this is the trace route:



1 2 ms 2 ms 1 ms 192.168.10.1
2 2 ms 3 ms 2 ms vpn-1.vm.***.net [192.168.10.8]
3 62 ms 61 ms 61 ms 10.1.0.1
4 71 ms 62 ms 61 ms project-vpc-1.vm.***.net [10.50.***.***]


I tried to lower the MTU to 1400 (1472 should be the best value for me) but nothing changed.

Update: I used ping -f -l xxxx to find the perfect MTU for my VPC destination. MTU perfect value is 1372. I set this value but I still have connection drops, I haven't fix this yet

Useful link (not so useful for me): http://www.heitorlessa.com/troubleshooting-ssh-connectivity-issues/

Update 2: partially solved, adding a permanent root (my sysadmin trick). I still have no idea why linux os works without adding that route and I have to add it, maybe the vpcvpn configuration could be involved [ http://blog.akquinet.de/2011/11/11/connecting-to-amazon-vpc/ ]