Have a border router running full tables and gigabit connections too couple BGP peers for likely ~9 months now. It replaced an Atom based Mikrotik. Router also does some basic firewalling and etc. Updated too latest Mikrotik release(6.19) week ago on it. Not really having any issues with it. BGP does seem to be single threaded so one CPU is maxed out frequently. Also have couple CCR's running 6.17 and terminating large number of PPPoE users. Have had issues with them but not for while. Not sure which release fixed it. Plan on updating them to latest Mikrotik release some evening just have not gotten around too it. I just don't think I would want to do BGP and PPPoE on same box myself.
On Wed, Sep 24, 2014 at 5:36 PM, Chris Wright via Af <[email protected]> wrote: > CCR-1036 running RouterOS 6.19 > > > > After some serious amounts of testing, we felt our CCR was ready to take the > plunge. The core router talks BGP to our two Imagestream Edge routers and > gets all 500k+ routes from each in about three minutes. Its PPPoE server > manages to authenticate the bulk of nearly 1800 customers in four minutes. > All’s fine and dandy for about 12 hours, then not so fine and dandy things > start happening. Overall traffic that should be near 600mbps seems to top > off around 400mbps. Edge 1 goes unresponsive, VRRP doesn’t kick in on Edge 2 > and the entire network degrades. All devices on our public switch go > partially unresponsive to pings including our DNS servers, other various > VM’s, and ESXi hosts themselves. > > > > Here’s the fun part: We took the CCR out, just flat out unplugged it and > turned on our old Core routers. They start authenticating customers but > they’re insanely slow in doing it. It’s not until we reboot our Edge 1 > router that things get back to normal and the old Core routers authenticate > at acceptable speeds. Could the CCR be inducing a problem in our Edge > routers perhaps? > > > > > > Chris Wright > > Velociter Wireless > >
