Hello,
Is it looping with the following network diagram, See the pic please. Note:Server A and Server B without routing function.
Don’t see a problem in this. As far as I know, “servers” do not participate in any bridging and will not be passing frames across 2 different NIC’s.
Thx reply, If three replace device replace with three routeros without bridging, should one of the links be blocked.
It is generaly not necessary to block one of links, depends on routing configured.
Generally routing loops only occur if the routing setup is somehow flawed. With basic routing made simple chances for routing loops are low.
In your example: if RouterA has default route pointing at cisco and only some specific route (e.g. 172.20.1.0/24) over RouterB … and similarly RouterB has default GW pointing at cisco as well, then packets will hardly circle around the three routers.
OTOH one can have a routing loop with only 2 routers involved in case when one is default gateway for the other one. The first one is configured to route traffic for some subnet towards second one and second one doesn’t have specific route for the same subnet and thus routing ut towards own default gateway (= back to upstream).
Routing loops are not very common. The diagram certainly is no routing loop. Actually the loop is in the routing table, not in the diagram. If there is no routing table (servers without routing) they cannot cause routing loops. Even with a routing loop in the tables the loop/path will stop with an “ICMP redirect” if the packet returns on the same interface. (To allow a packet to go in and out the same interface there must be a “spilt horizon” on that interface.) Dynamic routing tables (OSPF, RIP, EIGRP, …) will calculate the best routing for that instant. The IP routing is designed to work with redundant connections, it was basically the reason for creating the internet protocol, avoiding SPOF’s in times of war.
Ok, thx your replies.
Actually I intend to connect serverA and ServerB with straight cable without switch(10G network). It use to synchronize files between two servers (about 80TB hardisk)
192.168.1.0 only have 1G speed
Sure “multihomed” servers are still a common practice. Your example on 10G links direct between server is certainly possible and very fast for this purpose.
In this case you can create simple network of 2 devices, set both interfaces with appropriate IP addresses[*] and without any additional routes - this way the direct link will only be used for server2server communications.
[*] On linux, you can run
ifconfig <nic> 172.16.1.1 netmask 255.255.255.252 up
and use the …2 address on the second server. The trick is in the “long” netmask making this interface only suitable for server2server comunications. If you “wanted” a routing loop, you’d have to create it by adding more routes (which you probably won’t).
After that you have two ways if inter-server communications. If you use the 172… addresses, comms will go direct. If you use 192… addresses, comms will use “main” LAN via switches/cisco/whatnot .. and both ways can work in parallel at the same time.