Networking oddness

[uberlad]

This problem has left me and our IT boys shrugging our shoulders, so I'll see if anyone here has any ideas..

I have a shiny new webserver sitting on the DMZ at work. It serves webpages very nicely thank you. The problem is remote administration.

I'm connecting to it with RDP which was perfect when it was sat next to me inside the network. I could leave it logged in, come back the next morning and just have to log in. However, now it is in the DMZ after 10 minutes of inaction (or so, it's not a precise value) I come back to the RDP window to the message:

"The connection to the remote machine was broken. This may have been caused by a network error. Please try conecting to the remote computer again"

And then I have to connect again. Occasionally it won't even pick up the previous session so I have two sessions sitting on the remote machine, the active one not containing all the windows I opened. If this one dies then I get a choice of which one to connect to in future connections (max two connections set on the server)

All ports are open on the firewall between from LAN to DMZ, and RDPs between DMZ machines are not affected by this problem. The firewall doesn't appear to log anything, and adding a rule to allow RDP the other way though to my machine doesn't help with the problem.

Anyone have any idea what it is? Does RDP not like firewalls? Should I just give up on it and use something else for remote admin?

Thanks guys,

Uberlad

[edit - made little sense]

[VJ]

Have you tried forwarding the RDP port through the firewall?
i.e.
firewall port xx forwards to webserver port yy

On your RDP client, you then connect to firewall:xx



Jörg

[uberlad]

The rules on the firewall currently state 'Let all ports on the internal network straight through on to the DMZ network'.

Also I can RDP to my machine at home through the firewall without seeing this problem. I think (I'll have to check that).

I'm told by It that a new firewall is going online later, I'll test with that one to see if the problem persists.

Uberlad

[VJ]

Well, we have no DMZ set here, but our firewall blocks all incoming ports.

If we want to be able to connect via RDP to a computer on the internal network, we have to ask a portnumber on the firewall. The sysadmin then forwards this portnumber to the machine we like.

As a result, an RDP to firewall:xx directs you to a different computer than an RDP to firewall:yy. (no computers from the internal network are visible to the outside; all connections must pass the firewall)


Jörg

[uberlad]

Indeed. I've had to do the same thing at home to get RDP tunneled through my smoothwall. Hopefully the new firewall at work will just fix the problem.

Uberlad

[Byock]

I would say it is the Firewall. I had that problem before, and had to terminal in again, go to Terminal services manager, and reconnect to the existing session to see what I was working on. Changing to an OpenBSD firewall solved the issue for me.

[Wombat]

Are you running NAT or anything? My firewall/router reuses "idle" ports, so unless I have some form of keepalive going, I'll lose a connection after a while.

[uberlad]

Aha, NAT! I hadn't considered the fact that the port was being reused. This would fit with the situation.

I shall have to see IT in the morning and see if they can do anythgin about that port on that machine getting reassigned after long inaction. If not then I'll have to see if there's some way of forcing keepalives of some sort to be constantly sent between the machines.

I might even check the homepage of the firewall vendor, and see if they have it as a known problem.

Updates in 14 hours when I can get back in the office.

Uberlad

[uberlad]

Ok, update a little later than 14 hrs

I had to read the manual for the firewall but I found the conection timeout setting for the RDP rule and increasing it seems to have fixed the problem.

Which is nice.

Uberlad