Skip to content
Register

First full table from my basement router

Networking by LichunLars 23 replies 3.5K views
GuruDeutschlag

Member

#11
LichunLars said:
Pricing for even a 1G peering port makes me want to keep the tunnel forever

Ja this is the tragedy of the European internet infrastructure nein do not give up ja you should contact Netnod directly for their remote peering service nein it is not cheap but ja you can sometimes get a academic or research discount if you have the right affiliation nein I know a engineer in Frankfurt who pays less for 10G at DE-CIX than for his apartment ja but he works at a university

Emre10

Member

#12

Netnod pricing is a fantasy for individuals. I've looked at cross-connects from Istanbul to Frankfurt -- even with Turk Telekom wholesale, you're talking thousands of EUR monthly before you announce a single prefix.

Tunnels are the reality for anyone not running actual commercial traffic. I use Contabo's Munich location for my own BIRD instance, full table via a GRE tunnel to a friend in Amsterdam who has the real peering. 12ms added, but the monthly cost is a coffee.

LichunLars

Member OP

8 #13
Emre10 said:
Full table via a GRE tunnel to a friend in Amsterdam

This is exactly my setup except I'm the friend in Stockholm with the tunnel. Respect to the Amsterdam side -- that's where the real work happens.

How stable is your GRE? I see occasional flaps during Contabo maintenance windows, maybe twice a month. Always at 3 AM German time, which is civilized at least.

1ms or I don't want it
mediaaustin

Member

Deliverability Nerd

#14
LichunLars said:
This is exactly my setup except I'm the friend in Stockholm with the tunnel.

Fixed, thanks. Added the missing `+a` and `+mx` mechanisms. The domain is just for homelab nonsense anyway -- internal monitoring alerts and a private Matrix server -- but you're right that broken SPF is broken SPF.

Still amazed you spotted that from a forum signature. Deliverability nerd indeed.

SPF, DKIM, DMARC — holy trinity ✉️
tomchan

Member

#15
mediaaustin said:
The domain is just for homelab nonsense anyway -- internal monitoring alerts and a private Matrix server -- but you're right that broken...

This. I learned this the hard way with my shed setup -- Exim on the same Proxmox host as a Quagga test instance. Every time I flapped a session, mail delivery jittered enough to trigger monitoring alerts.

Now the mail VM lives on a separate NUC with its own UPS. Overkill? My wife's pottery newsletter subscribers disagree.

works on my bench ¯\_(ツ)_/¯
glasgow_tom

Member

#16

Following this with interest. My CS 1.6 server days taught me that latency matters more than bandwidth, but I never had to think about 935k prefixes. The whole thread is making me want to learn BGP properly instead of just renting game servers.

What's a good starting point for someone who understands networking but has never touched a full table?

LichunLars

Member OP

#17
glasgow_tom said:
What's a good starting point

Lab it with GNS3 or EVE-NG first. You don't need real hardware to understand path selection and communities. I ran fake full tables for six months before I touched production BIRD.

The real education is reading the RIPE mailing lists and watching the chaos during a route leak. You learn what actually matters when Facebook goes dark because someone fat-fingered a max-length announcement — https://bgp.tools

1ms or I don't want it
techmei

Member

#18

I run WireGuard overlays on everything, including my BIRD instances. The full table lives in a VPS in Taipei, tunnels to my apartment router, which only carries a default. No local BIRD, no basement heat, no cable management shame.

Latency to Cloudflare is 2ms from the VPS, 5ms through the tunnel. Good enough for my use case. The idea of running 935k prefixes on local hardware feels like a hobby for people who enjoy suffering.

SamAlvi

Member

Self-Host Everything

#19
techmei said:
The idea of running 935k prefixes on local hardware feels like a hobby for people who enjoy suffering

Some of us enjoy the suffering. Also, local BIRD means I can play with communities and prepends without worrying about a VPS provider's acceptable use policy. I've had hosts threaten to cancel for "excessive BGP sessions" before.

Your WireGuard approach is clean though. I use it for management, not data plane. Different tradeoffs.

my cloud. my rules. my 3AM alerts.
RA

Member

#20

Has anyone actually gotten a full table from RackNerd? I have a VPS in their Los Angeles location, but when I asked about BGP they said "no full tables, default route only." This was last month.

Wondering if it's location-dependent or if I got a support rep who didn't know the policy.

Post a reply

You need an account to reply. Log in or register to join the conversation.

Post reply Preview Save draft