We've recently changed VPNs from OpenVPN to SoftEther; it's working great so far, with one exception: SQL Server Management Studio. For some reason, either SSMS or the SQL Server seems to drop the connection after a short amount of time (~30 seconds). Trying to run a query after that (even a simple SELECT) returns this:
Hitting F5 at that point re-establishes the connection and everything works OK...for another 30 seconds or so. Not a deal-breaker, but massively annoying if you're a DBA (like several of my users).
Everything else works fine over SoftEther: file + print, SSH/RDP sessions, MySQL Workbench, whatever. It just seems to be SSMS that doesn't like it, but I don't know why. The issue is common across all SQL Servers that I've tested it with; I've adjusted timeout/keepalive settings on the server and client, but nothing seems to have any effect. A packet trace of SQL Server traffic shows the server sending a RST after about 30 seconds:
(192.168.30.30 is my VPN IP, 172.16.1.132 is the SQL Server I'm testing with)
Does anyone have any idea how I might persuade this thing to stay connected? Angry DBAs are no fun to be around.
EDIT, in response to Greg's comment: traffic on the server shows no RST, just a FIN/ACK:
(150.150.10.39 is the LAN-side IP of the VPN server, which is NAT-ing my traffic)
Further updates, 14/3/22: Adjusting the MTU on the client, VPN server or SQL server appears to have no effect. What does have an effect is using DBeaver as the client instead of SSMS; that's rock-solid, with no timeouts or disconnections.
