I'm working on setting up an anycasted service with the containers hosted on Proxmox. The current plan is to give each container a separate IP address in the same subnet as the router via adding a regular network device in Proxmox. Within each container they'll be configured to add the anycasted IP address to a loopback interface. That won't be a network device under the network tab in Proxmox, just entirely managed and assigned within the container itself. Is this safe to do if the containers get migrated to the same host during e.g. some maintenance or failover? Here's how I want to configure it.
router: 10.0.0.1/24
CT1: eth0 10.0.0.2/24
lo 10.10.10.10/32
CT2: eth0 10.0.0.3/24
lo 10.10.10.10/32
And just run ExaBGP within the container to advertise or withdraw the route from the router as needed. I'm assuming that each container being in their own network namespace ought to make this work fine, but the duplicate ip addresses in the same kernel just seems weird to me. Is this a sane plan?
router: 10.0.0.1/24
CT1: eth0 10.0.0.2/24
lo 10.10.10.10/32
CT2: eth0 10.0.0.3/24
lo 10.10.10.10/32
And just run ExaBGP within the container to advertise or withdraw the route from the router as needed. I'm assuming that each container being in their own network namespace ought to make this work fine, but the duplicate ip addresses in the same kernel just seems weird to me. Is this a sane plan?
