Tue Mar 09, 2010 9:16 pm
I had the same problem and had a support ticket open with MT support. I think I figured it out, though. try changing the profile to "only-one=no"
I think the server was colliding with itself when assigning IPs to clients if they disconnected and reconnected before openvpn realized they were gone. Since I changed this, I havent had a lockup. I was having it happen at LEAST weekly, normally ever 2-3 days, and sometimes twice/day. Right now, I am at 10 days since the last restart and that restart was an accident (didnt pay attention to what router I was in). So it has been 14ish days since I have seen it happen.
I use 2048bit certs for every client, require client cert, clients are a /32 point to point link, server address for every client is the same. AES-128 w/ SHA.
Here is the profile I use for 82 of the 86 connections:
/ppp profile
add change-tcp-mss=no comment="" local-address=10.16.128.1 name=Customer \
only-one=no use-compression=no use-encryption=required \
use-vj-compression=no
/interface ovpn-server server
/interface ovpn-server server
set auth=sha1 certificate=fw-1 cipher=aes128,aes192,aes256 default-profile=\
Customer enabled=yes keepalive-timeout=60 mac-address=FE:1D:10:11:78:AB \
max-mtu=1500 mode=ip netmask=32 port=1194 require-client-certificate=yes
clients are authenticated via RADIUS to a FreeRADIUS server w/ MySQL backend behind the RB1000.
Like I said, Im at almost 10 days of uptime right now and ~100gb of traffic has traversed the firewall. RouterOS v4.5