I have heard a lot of bad about the BGP and OSPF in MT.
A lot of it has centered around bugs in basic functionality as well as lack of featuers.
Is it to the point where basic BGP and OSPF functionality works? If so, how far does it extend from that? If not, what doesn’t work and how?
From what I have seen on the forums I think it distills down to “Don’t use it in production”.
However I have not personally used it, at all.
I have not tried it, but I’m guessing that setting up my own Linux with Zebra or whatever the daemon of choice is these days would be difficult. Does anyone else have software that reliably does BGP and OSPF to load onto x86 or x64 hardware?
I dont know about bgp but ospf work flawless if it’s setup ok with a few workarounds (using bridges, wds etc)
please elaborate. bridges? then why route?
Hello
This is what I do…
wlan1 added to bridge1 & wds dynamic bridge1 192.168.0.1(assigned to bridge)
wlan2 added to bridge2 & wds dynamic bridge1 192.168.1.1(asridged to bridge)
So really I use bridges but wlan and wlan2 remain not bridged
I have never had a ospf problem since then, no init state, no ex-start, no problems
I have no idea why this makes ospf more stable, I think/guess its because a bridge is never disconnected
So then you lose one of the main reasons of a dynamic protocol… it senses unusable interfaces and routes around them.
OSPF will still drop the route after 40 seconds (unless default is changed) of no-communication with the peer, so it’s not completly useless. it just isn’t a transparent failover.
At least it works…
And yes, you could sdjust teh 40 seconds
I would lik eit fixed so I would not need my workarond but untill then this is fine.
Other wirless platforms seem to struggle with ospf also I have noticed
ospf is actually a very complex protocol to implement and is often implemented as a state engine (no sessions to help it keep track of who’s on and who’s off )
ospf needs to send/receive periodic updates to know partner state.
ospf is a link state protocol and wireless’ idea of link state is a little … obscure.
most implementations are based on zebra (often via quagga) and zebra was fundamentally flawed.
bgp does not suffer from the protocol specific issues but it’s quite complex as well. rip… well uh, it works … i think. i dont touch dynamic routing protocols unless i have to, i love bgp but since anything dynamic is (more often than not) the cause of less uptime.
2.9.27 ospf routing-test works quiet fine. Anythink after that is broken more or less (dont know about last two distros, as they have some other bugs). That a 4 months of broken software, and no way to download 2.9.27 if someone does not have it - LOL.
any news about fixing OSPF?
thank you
our entire network is ospf, there are only a select few static routes. it is generally very stable, from v2.9.10 through 2.9.28 (but not 15, 17 and 27, had significant issues with those builds) once in a while a system will just break ospf, and need rebooted, we have over 50 devices running MT, that only happens once every 3-4 weeks. Also, after disableing an interface, and re-enableing it, OSPF will many times not turn back on (just shows a state of INIT).
Haven’t tried BGP yet… we’re going to take one of our providers off of our Cisco 7204 (currently has eBGP sessions with our 2 upstream providers) and move them on to a MT box, that will be running iBGP with our Cisco 7204, so we won’t have a single point of failure at the 7204.
When we get closer to implementing that, I plan on getting a buch of peoples opnion on the best/most stable build to do that.
Hello
first of all thank you for your reply. I think we will try to implement OSPF on our network. So I need some more info:
Thank you!
regular routing package… I won’t use a test / beta package in production, that’s just asking for trouble.
we’ve got one system running 2.9.31, seems fine, but as of yet, I’ve got no reason worth the potential problems or effort to move past 2.9.28.