🌗 自建自治系統：在FreeBSD上使用FRR、GRE隧道與策略路由實現BGP
➤ 從零打造個人化全球可路由網路架構
✤ https://blog.hofstede.it/running-your-own-as-bgp-on-freebsd-with-frr-gre-tunnels-and-policy-routing/
本文完整解析如何透過贊助LIR取得AS號碼與IPv6位址段，在FreeBSD系統架設BGP路由器，結合FRR路由套件與GRE/GIF隧道技術，實現個人自治系統運作。作者詳細說明從資源申請、網路配置到策略路由的實作流程，特別針對伺服器需同時使用多組IPv6位址的技術難題，提出透過隧道分發子網段的解決方案。文中更包含實際配置範例與網路架構圖解，展現如何用單臺虛擬機將自有IP位址宣告至全球路由表。
+ 原來個人也能申請AS號碼，這篇實戰指南完全打破我對BGP的傳統認知
+ 策略路由與隧道配置的細節說明非常實用，正好解決我跨機房流量分配的問題
##FreeBSD #BGP #網路架構 #IPv6 #FRR
#FRR
IPv6 should be treated the same as, if not better than, IPv4!
https://github.com/FRRouting/frr/issues/20067
https://github.com/FRRouting/frr/issues/20102
containerlab × FRRoutingで実現！BGP冗長化シナリオと自動経路切り替え 実験レシピ
https://qiita.com/ntaka329/items/1a4291591dc63a55a05c?utm_campaign=popular_items&utm_medium=feed&utm_source=popular_items
お家でBGP！containerlab x FRRoutingで学ぶBGP経路制御シミュレーション
https://qiita.com/ntaka329/items/22a9f1ad999396b76b0a?utm_campaign=popular_items&utm_medium=feed&utm_source=popular_items
【検証】MacBook M4でContainerlabを動かしてみた！ARM対応NOSでネットワーク検証環境構築
https://qiita.com/ntaka329/items/ed5d8b0b9de27893e0a7?utm_campaign=popular_items&utm_medium=feed&utm_source=popular_items
and here the RFO for our 5-10 minute of problems today:
Oh look, it's an "Ubuntu 24.04 pushed out a systemd-networkd update and broke things" week again.
Did you know that #systemd-networkd *really* dislikes it when anything else tries to manage the system's route table, and when it restarts it will frequently remove a bunch of routes that were added by dynamic routing, #K8S CNI, or other similar tools?
This has caused a number of fairly public outages for sites over the past year or two. There are a set of config flags for /etc/systemd/networkd.conf that will disable this behavior, but the last of them went in in Systemd v256, and #Ubuntu 24.04 (LTS) uses v255.
I have a writeup with some workarounds: https://scottstuff.net/posts/2025/02/25/frr-vs-systemd-networkd/
In my case, I tend to lose a bunch of either IPv4 or IPv6 (but rarely both) routes when `networkd` updates roll out on their own. My first indication of trouble is usually a ping-check alert firing.
My specific problem is caused by #FRR; it defaults to creating its own kernel next-hop groups for routes learned from OSPF or BGP, and `systemd-networkd` then comes along and cleans them up, which causes the kernel to drop all of the routes that were associated with them. FRR fails to re-add the routes until it's restarted, so things stay broken until they're manually cleaned up.
I've been testing a fix on ~half of my machines since July to see if it caused any *other* problems, and it seems to have passed. Adding `no zebra nexthop kernel enable` to FRR seems to avoid the nexthop problem without causing any other issues for me. This *may* break if you use policy-based routing with FRR, but it's probably safe otherwise.
L’EUROPE INVESTIT AUTREMENT
Prêts, garanties, fonds propres : les instruments financiers du plan Next Generation EU prolongent l’impact au-delà de 2026. Une Europe qui prête pour bâtir, plutôt que subventionner sans lendemain.
https://whatsapp.com/channel/0029VaE5Wl8Dp2Q7vJyg0G30
#EuroScope #NextGenerationEU #Relance #Investissement #TransitionVerte #FRR
Just made a small app for Flammenrouge racing tour :)
Анонсируем Full-View на изолированном стенде
Мы разрабатываем сервис, который поможет сетевикам управлять аплинками. В процессе подготовки стенда для тестирования потребовалось в изолированном окружении развернуть виртуальный маршрутизатор и анонсировать на него несколько Full-View.
FRRouting. Без кликбейтного названия
Вот есть у нас, например, два хоста-клиента и есть один Linux. Хосты из одной сети. Надо чтобы пообщаться смогли они друг с дружкой...
#frr weirdness: ospf6d accumulated a ton of duplicate routes for 2 peering links and an anycast subnet around 66-68 days ago. there's 13.4k entries for the anycasted /64 and 5.6k/5.8k for the two /127s
none of the other routes are unusual
@rachel I have been in my software router era for a very long time with #VyOS then #openwrt then back to #VyOS tried to sprinkle #frr around there too. It just took a little bit of cognitive load when updating the host or the software I'm using as my virtualization layer.. Now! I just been on this router for like 4 days.. I might be full of shit 😂😂😂
I am confused by FRR.
https://docs.frrouting.org/en/latest/bgp.html
"-l, --listenon - Specify specific IP addresses for bgpd to listen on[...] Note that this option implies the –no_kernel option, and no learned routes will be installed into the linux kernel."
But they are installed into the kernel table! I can see it (with ip -6 route show table 50).
Is this because it's an bgpd option, and the routes go via zebra instead?
Config:
https://privatebin.net/?6461c22964ba20b8#BtwSKdrrRPu2tx8DQwfqyBaDU2gjdD5MqSYTUejXBTMq
and
https://privatebin.net/?34bd84b23b49a596#AVw68P7USz458c8hgk91GZ9dBmJZCADEw2Mq2gYtBw7y
And even if it did. what if I want to run multiple instances of bgpd and to install the routes into the kernel?
edit:
The docs seem to be wrong. Adding "-n", the routes are indeed not installed in the routing table.
Moved my VPS from #Rocky 9 to #NetBSD 10.1. Surprisingly it works fine - sans a quirky GRE implementation... I made it work in the end but all other network devices are able to use any subnet on a GRE interface.
In the end I have:
- #wireguard to multiple #MikroTik routers
- GRE tunnels to all of the routers
- #FRR running #BGP
- #haproxy doing reverse proxy for #Nextcloud and other services
- #certbot renewing Let's Encrypt certificate
I got tired of fussing with #FRR and trying to use it to do VRF's and routing as it really is geared towards Dynamic Routing! Back to #VyOS for the #homelab. I am still using VRRP with 3 FRR routers on my #incus hosts. So as long as my upstream router is up. I can reboot hosts all day long and no internet go down.
Yo!! this shit kinda 🔥!! I'm using 2 lxc containers that are using the FRR project. I configured the #frr packages to use #VRRP for both lxc instances. I now have redundant gateways for the
#homelab . I dropped one packet as it switch over!
VyOS 1.5 is somehow broken for me: Babel does not work at all (not even with the example from the documentation) and with BGP you cannot set a source address for installed routes (possible with route-map set src, but has no effect).
The rolling release seems to be enormously rolling.
Client Info
Server: https://mastodon.social
Version: 2025.07
Repository: https://github.com/cyevgeniy/lmst