zebra server redundancy
A while back there was a thread about using zebra/gated to originate a host route into OSPF that is tied to the loopback on multiple boxes, in order to achieve server redundancy and load balancing w/o appliances. In testing so far, this seems to work great. My only question is how are people withdrawing the routes? ospfd does not yet re-read its config on a HUP. Killing the daemon seems a bit rude, depends on timers for convergence, and precludes having more than one service per box. removing the alias from the loopback interface seems slow to converge, as zebra does not seem to be notified in real time of such events, but seems to have a periodic timer that looks at interface status. (Adding it back to the loopback has also had some issue, if the host route is already in the routing table from other hosts.) Looks like an expect script telneting to the ospfd and removing the network statement from the router ospf section is the best way, or is there some more elegant mechanism people use? Thanks
This isn't qiute a nanog subject, but...
In testing so far, this seems to work great. My only question is how are people withdrawing the routes? ospfd does not yet re-read its config on a HUP. Using vtysh (or telnet to port 2604)
removing the alias from the loopback interface seems slow to converge, as zebra does not seem to be notified in real time of such events, but seems to have a periodic timer that looks at interface status. (Adding it back to the loopback has also had some issue, if the host route is already in the routing table from other hosts.) Depends on your OS. zebra will get realtime notification on Linux with netlink, and on *bsd with "routing socket".
Looks like an expect script telneting to the ospfd and removing the network statement from the router ospf section is the best way, or is there some more elegant mechanism people use? Thanks Vtysh is somewhat simpler than telnetting.
Alex Pilosov | DSL, Colocation, Hosting Services President | alex@pilosoft.com (800) 710-7031 Pilosoft, Inc. | http://www.pilosoft.com
Depends on your OS. zebra will get realtime notification on Linux with netlink, and on *bsd with "routing socket".
And then there are times when netlink communication dies and causes a bgp prefix to get 'stuck' as kernel route :( hahah Never had problems with socket/ioctl though. -hc -- Sincerely, Haesu C. TowardEX Technologies, Inc. WWW: http://www.towardex.com E-mail: haesu@towardex.com Cell: (978) 394-2867
Looks like an expect script telneting to the ospfd and removing the network statement from the router ospf section is the best way, or is there some more elegant mechanism people use? Thanks Vtysh is somewhat simpler than telnetting.
Alex Pilosov | DSL, Colocation, Hosting Services President | alex@pilosoft.com (800) 710-7031 Pilosoft, Inc. | http://www.pilosoft.com
participants (3)
-
alex＠pilosoft.com -
Haesu -
Steve Francis