[Babel-users] High CPU usage & instability with full IPv6 table

[Steven Honson]

Thank you for such a prompt reply Toke!

On Fri, 10 Jun 2022, at 11:34 PM, Toke Høiland-Jørgensen wrote:
>> At the moment I am using Bird 2 to handle both BGP and Babel, and it
>> barely breaks a sweat, but would like to use babeld for the Babel
>> component so to be able to leverage some of the additional features
>> that the Bird implementation lacks.
>
> Which feature(s) are you missing in Bird? :)

The two that I'd like to use are IPv4 with IPv6 next-hop (RFC 9229), mainly to keep my network as IPv6 single stack as possible, with just IPv4 mainly on loopbacks, and delay based metrics (draft-jonglez-babel-rtt-extension-02), as in my case some Internet routing quirks mean that least-hops isn't always the lowest latency path between my hosts.

Neither are critical, and bird is doing a wonderful job, so it isn't a big issue if I just keep forgoing these.

>> Any clues what sort of limit I may have reached or how to mitigate
>> this? Is the CPU being eaten up constantly monitoring and filtering
>> route changes in my kernel routing table due to the constant churn of
>> the global IPv6 table?
>
> Sounds like you could be hitting the same issue as described in this
> thread?
>
> https://alioth-lists.debian.net/pipermail/babel-users/2022-February/003884.html

That does indeed look like the same issue, thank you for the reference, I've been subscribed to this list for quite a while now and somehow missed that. That helps me understand what is going on a lot better.

Cheers,
Steven