Daniel Lezcano <[EMAIL PROTECTED]> writes: >>>By the way, having the same IP address for several containers, how will be >>>possible to do container migration ? >> It depends on the circumstances. In general having several containers with > the >> same IP address is a bad idea. But if you have a setup where you can >> do it safely there is nothing preventing that setup from working between >> machines, so it is neither a positive or a negative from a migration > standpoint. > > What about the clients connected to the different containers ? For example, > you > have 100 containers, all are configured with 192.168.1.100 and have an > application binded to INADDR_ANY:80. > In front of that, you have 100 clients. Each of them are connected to the > application running in each container. > One container is migrated to another machine on the network, it is not > possible > to keep the same address, so the address is changed to 192.168.1.200. The > clients connected to 192.168.1.100 will lose the communication.
So don't set your network up so that if you move a machine it's IP address must change. Either that or you require everyone with 192.168.1.100 to migrate at once. You might need a tunnel over the network to make the configuration you describe work. This is a policy issue on top of a basic mechanism. I do not encourage anyone to reuse the same IP address on multiple containers and setting up the routing is a tricky proposition. But it is exactly the same problem as using the same IP address on multiple machines. So weather you use real network devices or logical ones does really matter. I think the polyinstantiated bind that people were discussing earlier is suspicious for exactly the same reasons. Multiple uses of the same IP address relies on network magic to make it work. Network magic with out a lot of care is fragile. At the same time there are legitimate uses of hosts with the same IP address such as host failover, in a high reliability configuration. Eric -- redhat-lspp mailing list [email protected] https://www.redhat.com/mailman/listinfo/redhat-lspp
