Going crazy here, desperately looking for some help. We have 5 terminal sessions, with users using thin clients to RDP to them. With Cisco 3560g switches, we never had a problem. We switched to Meraki switches, and brief disconnects occur at random times. The users are booted from session, and when logged back in, the session continues right where they left off. We noticed our physical hosts (where the TS reside) had MTU set to 1500, so we changed Meraki MTU settings to match, at 1500. This stopped probably 80% of disconnects. But can't seem to find how to make the disconnects completely go away. I see some TCP fragmentation coming from terminal server to client, with packet size as 1518. Perhaps i need to raise the MTU on MEraki side to 1520 to account for header?
No CRC errors that we can see. We changed uplink cables, SFP connectors, and everything else we can think of.
You should not be needing to mess with MTU settings for RDP servers in a local LAN environment,
It's not beyond the realm of possibility as an issue, as the 3560's used a 1500 byte MTU, while all Meraki switches support jumbo frames (around 9000-byte MTUs).
The MTU is negotiated between the client and the server and the smallest of the two is selected. It is highly unlikely your clients have an MTU of anything other than 1500. So configuring a higher MTU on the server is likely to have zero impact. Configuring a lower MTU will result in a small MTU overall, but there is not a high probably of this altering your problem.
The only recent change you mention was the switches. Have you double-checked the port connections to the server are operating at full-duplex, and that there are no Ethernet errors to speak of? Ideally, configure everything for auto/auto.
Have you double-checked the connections to the clients, to make sure they are auto/auto and negotiating a full-duplex connection, and that there are no Ethernet errors reporting on the MS switch port?
Anything in the Meraki event log for switches? Any spanning tree issues reported, port problems, or anything else interesting? Duplicate IP address warnings?
If you test it off one of your thin client devices and a Windows computer - are both clients equally affected?
Are there any firewalls in between the users and the servers? If so, is it reporting anything interesting?
If all else fails, you are going to need to take some packet captures at the server and client end to understand what they are seeing.
I'm running a full catalyst switching stack, with Meraki MX interconnecting each location using the Autovpn, ever since implementing the Meraki's with Auto VPN my RDP connections are having connection issues. I perfomed wiresharks and the packets is the wireshark look to be at 1240 to 1400 in size. I'm tempted to manaully redice the MTU on my test devices to see if it makes a difference, but at least per wireshark MTU shouldn't be a problem. I'll update soon.