Take the 2-minute tour ×
Super User is a question and answer site for computer enthusiasts and power users. It's 100% free, no registration required.

I found some similar questions, but not quite like this, so am going ahead and posting.

I use a replication product at a client site. We are running a Windows 2003/2008 server environment. It is not working very well because it claims that it's losing connection with its agent(s) from time to time. But it does this even on LAN connections, and being on a very solid corporate LAN, this seems highly unlikely.

However, getting Support to agree that it may be their issue and not ours, is where I need help. I need to gather some kind of evidence that another program that relies on a steady TCP connection does not report drops when the replication product does.

So I'm looking for a utility that would run on 2 servers, and can establish a TCP connection between them, and then report back on any connection interruptions that ensue, even if brief. It doesn't even have to be a utility, but can even be a program that relies on something on the other server so much that if there was even a momentary drop, it would display a message or error. I'm pretty flexible, but I just need a reliable way to prove that the connection is solid.

I thought about continuous ping, but its limitation is that it has no timestamp associated with any dropped pings, and is also not using TCP, so may be a slightly different scenario.

Thanks much in advance! -Sandra

share|improve this question

1 Answer 1

A packet capture is your fried here. Start a long running capture open with a filter for just the two servers set, also run it on both servers to compare the results. When troubleshooting something like this I look for resets on the connections and any long waits between connections.

Wireshark will do the trick. Set the capture filter on Server1 to 'host server2' on Server2 set the capture filter to 'host server1'. This will capture all traffic between the two. You may also want to set the Limit each packet to 128 bytes so it does not grow to big.

Also for your idea of a heartbeat tool. http://www.codeproject.com/Articles/10481/Simple-ICMP-Service-Monitor

share|improve this answer
    
If the application can log, make sure you have logging set as high as it can go (both server and client). Then employ packet captures (on both server and client). Wireshark will let you do them in a rolling fashion so that you don't use up all your HDD space, but can leave it running. Then monitor for errors from the application. Once you get an error, stop the packet capture and pull the logs. Then look at timestamps for errors in logs and find those timestamps in pcaps and look for anything out of the ordinary. –  MaQleod Jul 7 at 15:40
    
Ok, I'm no network expert, so although examining traffic may be the ultimate solution, it's not the one for me. I'll try the heartbeat tool, although it's ICMP-based rather than TCP connection based, so not entirely what I was hoping for, but hopefully close enough. Thanks –  Sandra Jul 9 at 13:59
    
And yes, the replication tool has detailed logs, so I can know the exact times when it thinks it lost connection to its agent. –  Sandra Jul 9 at 14:00

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.