PetraSuper
Member
OP
network boot believer
- Joined:
- Jun 2024
- Posts:
- 127
- From:
- Bratislava, Slovakia
He, I write malý guide from morning when I catch own leak in 90 second
No, you need only three free thing:
- bgp.tools for quick prefix check
- route-views or looking glass from OVHcloud
- malý script what curl both and diff
On server I run bird with full table. Into config I have upstream from Contabo and Vultr. No, I set local-pref wrong on one session, 90 second later bgp.tools show my /48 coming from wrong path. Hej, I catch before any mail from peering partner.
No, key is check yourself before someone else do. I run script every 15 minute from cron. No fancy, just curl and grep.
You have similar setup? What tool you use for catch own mistake?
boot anything, anywhere, anytime
tomchan
Member
- Joined:
- Jul 2024
- Posts:
- 202
- From:
- Bristol, UK
Fun little project — I run similar on my NAT VPS from RackNerd, just a bash loop with bgpq4 and some looking glasses
#!/bin/bash
ASN="64496"
PREFIX="2001:db8::/32"
for lg in "OVHcloud" "Vultr" "Hetzner"; do
curl -s "https://${lg}.example/lg?cmd=show+route+${PREFIX}"
done | grep -v "${ASN}:$" && echo "LEAK DETECTED"
It actually works, caught my typo last month
works on my bench ¯\_(ツ)_/¯
PetraSuper
Member
OP
network boot believer
- Joined:
- Jun 2024
- Posts:
- 127
- From:
- Bratislava, Slovakia
Hej, full table on NAT VPS is crazy. I have real server in Bratislava, not VPS. Bird with full table need real RAM. But for only own prefix check you dont need full table, just looking glass from outside. My script is only curl https://bgp.tools, no bird on small box.
boot anything, anywhere, anytime
tomchan
Member
- Joined:
- Jul 2024
- Posts:
- 202
- From:
- Bristol, UK
Same here, dual check. I actually caught OVHcloud LG giving stale data once, route-views saved me from false panic. Always have two sources.
works on my bench ¯\_(ツ)_/¯
PetraSuper
Member
OP
network boot believer
- Joined:
- Jun 2024
- Posts:
- 127
- From:
- Bratislava, Slovakia
RIPE RIS is good but delay is bigger. For 90 second catch you need something faster. Bgp.tools update quick, route-views also. RIS is for later analysis, not for wake up at 3am.
boot anything, anywhere, anytime
