sofialund
Member
OP
homelab heatstroke
- Joined:
- Jul 2024
- Posts:
- 140
- From:
- Buenos Aires, Argentina
¡Hola all! I have the session new with Contabo and need to test before I announce the prefix live. I use the lab small in GNS3 but the router virtual eats all the RAM of the server red. ¡The topology is simple but the convergence slow!
I think in use the environment of production with prepends long and communities no-export but my boss say is too risky. ¡How do you make the test? I need know if the filters of they are correct!
¡Thank you!
hot air, steady hand, magic smoke
GeorgeNmp
Member
AS64512
- Joined:
- May 2024
- Posts:
- 218
- From:
- Ashburn, US
I think in use the environment of production with prepends long and communities no-export
This is actually more common than people admit. Proper staging for BGP sessions requires either:
- Physical lab with matching firmware versions and identical policy configurations
- Virtual environment (GNS3, EVE-NG, CML) with sufficient resources for control plane scale
- Route server testbed at an IX if you are peering publicly
- Production announcement with NO_EXPORT, well-known communities, and maximum prepends
The last option is functionally what many operators do for "soft launches." The difference between a lab and reality is always the full routing table, the actual neighbor policies, and the RPKI/IRR validation path. A virtual IOS-XR instance does not have the same memory behavior as an ASR 9000 with 1.2 million prefixes.
That said, if you do use production, you need monitoring in place and a withdrawal plan measured in seconds, not minutes. I check prefix propagation with https://bgp.tools during these tests.
iBGP, eBGP, don't care, just peer