MikroTik

Below you will find a chunk of the log from a mikrotik which is a running pppoe-server for an ISP with 750 users.

We are having trouble with peole losing their connections. The log represents what happens when people lose a connection. From the log, it looks as if the network cable gets unplugged.

The radius logs show lost-carrier with a session time of 25 seconds for the majority of the lost connections.

What we are trying to figure out is why people get logged on an then immediately get logged off.

From what we can tell, the connections are mostly from XP pppoe clients.

The mikrotik is running on a 1.8 Ghz AMD Athlon with 1Gb of RAM. The pppoe servers are configured on vlans and are dispersed over two RB44 cards to minimize the size of the broadcast domains. The mikrotik seldom runs over 50% CPU utilization. The uplink port is an intel e100 card.



5:43 pppoe,ppp,info,account dougami logged out, 95 5753 135233 97 148
19:45:43 pppoe,ppp,info <pppoe-dougami>: terminating... - peer is not responding
19:45:43 pppoe,ppp,info <pppoe-dougami>: disconnected
19:45:46 pppoe,ppp,info <pppoe-0>: terminating...
19:45:46 pppoe,ppp,info <pppoe-0>: disconnected
19:45:50 pppoe,ppp,info,account pebaker logged out, 695 257314 2283062 1638 2235
19:45:50 pppoe,ppp,info <pppoe-pebaker-1>: terminating... - peer is not responding
19:45:50 pppoe,ppp,info <pppoe-pebaker-1>: disconnected
19:45:50 pppoe,info PPPoE connection established from 00:18:F8:DE:F7:45
19:45:50 pppoe,ppp,info <pppoe-0>: waiting for call...
19:45:56 pppoe,ppp,info <pppoe-0>: terminating...
19:45:56 pppoe,ppp,info <pppoe-0>: disconnected
19:45:57 pppoe,ppp,info,account la2viburnum logged out, 638 301927 9814766 4206 6908
19:45:57 pppoe,ppp,info <pppoe-la2viburnum>: terminating... - peer is not responding
19:45:57 pppoe,ppp,info <pppoe-la2viburnum>: disconnected
19:45:58 pppoe,ppp,info,account porterat logged out, 515 731677 5230932 4129 5120
19:45:58 pppoe,ppp,info <pppoe-porterat>: terminating... - peer is not responding
19:45:58 pppoe,ppp,info <pppoe-porterat>: disconnected
19:46:00 pppoe,info PPPoE connection established from 00:E0:18:A5:91:01
19:46:00 pppoe,ppp,info <pppoe-0>: waiting for call...
19:46:00 pppoe,ppp,info <pppoe-0>: authenticated
19:46:00 pppoe,ppp,info <pppoe-0>: connected
19:46:00 pppoe,ppp,info,account estrellas logged in, 66.43.37.161
19:46:10 pppoe,ppp,info,account brenn logged out, 1775 2379986 20536765 11541 16738
19:46:10 pppoe,ppp,info <pppoe-brenn>: terminating... - peer is not responding
19:46:10 pppoe,ppp,info <pppoe-brenn>: disconnected
19:46:11 pppoe,info PPPoE connection established from 00:13:D4:22:9D:45
19:46:11 pppoe,ppp,info <pppoe-0>: waiting for call...
19:46:11 pppoe,ppp,info <pppoe-0>: authenticated
19:46:11 pppoe,ppp,info <pppoe-0>: connected
19:46:11 pppoe,ppp,info,account tgildig logged in, 66.43.37.168
19:46:12 pppoe,ppp,info,account foodoe logged out, 156 93133 922669 853 1112
19:46:12 pppoe,ppp,info <pppoe-foodoe>: terminating... - peer is not responding
19:46:12 pppoe,ppp,info <pppoe-foodoe>: disconnected
19:46:12 pppoe,info PPPoE connection established from 00:19:21:0F:05:F8
19:46:12 pppoe,ppp,info <pppoe-0>: waiting for call...
19:46:13 pppoe,ppp,info <pppoe-0>: authenticated
19:46:13 pppoe,ppp,info <pppoe-0>: connected
19:46:13 pppoe,ppp,info,account woodall logged in, 66.43.37.163
19:46:13 pppoe,ppp,info,account spidero logged out, 155 8905 168624 136 174
19:46:13 pppoe,ppp,info <pppoe-spidero>: terminating... - peer is not responding
19:46:13 pppoe,ppp,info <pppoe-spidero>: disconnected
19:46:13 pppoe,info PPPoE connection established from 00:11:11:B8:F0:B7
19:46:13 pppoe,ppp,info <pppoe-0>: waiting for call...
19:46:13 pppoe,ppp,info <pppoe-0>: authenticated
19:46:13 pppoe,ppp,info <pppoe-0>: connected

Turned-off keep-alives on the pppoe servers and fixed the auto disconnect problems. But, I now must live with the issues associated with the lack of keep-alives.

As well, we now have another problem. It seems that every once in a while, while we are under a pretty significant load, the pppoe servers just stop working. The first time, it just impacted a couple of servers. But, it eventually impacts all pppoe servers running on the MT.

We are trying to figure out if it is related to load. We have plenty of memory. But, we expect we are processor bound.

Any ideas would be helpful.

A quick and dirty rule of thumb....

Allow for 2mhz per user.

IE: we have 266 mhz boards that we run as AC's. When a site gets to 120 users we put in another board and run them in parallel - if we let the site get to 150 without adding an extra board then we see it start to slow down.

We also have a problem still where the PPPoE locks up if a significant number of users fall off/reconnect - but running 2+ boards in parallel has helped with this also.