We have four networks for four departments: R&D, Sales, admin and finance. All IP addressing is from the network. So, maybe four PCs have same IP address, e.g.

Now, we want to connect them together with a router, so our CEO may access some servers in different departments, which each may have same IP address now.

I knew how to set up NAT and internal servers for the router so outside networks can access the internal servers, but now the four networks have same IP addressing.

We can add multiple network cards to server, or we can move network addressing to a larger network, then we have enough IP addresses, but both of those solutions are not accepted by end user.

I already use and H3C router to set NAT and internal servers for different sub networks. but now the networks have same IP addressing.

Expected result:

  1. Connect four networks by a router
  2. Do not change the current four network IP addresses.
  3. The CEO want to access some servers inside different networks, each of which may have same IP address
  • 2
    Network classes are dead (please let them rest in peace), killed by RFCs 1517, 1518, and 1519, which defined CIDR (Classless Inter-Domain Routing) in 1993. Modern networking does not use classes. The easiest thing to do is to renumber three networks. There is no reason to have four networks with the same addressing, and it will cause difficult problems in the future. Routers route traffic between different networks, not from a network back to the same network. NAT is not a substitute for routing. You should probably ask this question on Network Engineering, not the SE programmer site. – Ron Maupin Jul 22 '19 at 0:52
  • 4
    The correct approach is to renumber 3 of the networks so the IPs do not overlap. You could use 192.168.[2,3,4].0. Or if there are fewer than 60 hosts per net you could use, and so on. NAT will not help because each host currently thinks it is directly attached to the subnet so they don’t even send packets to the router. I suppose you could give the CEO 4 wires, on on each net, and have him change cables as needed. But plugging in 2 cables at once won’t help because the computer won’t know which interface to send the packets. – Darrell Root Jul 22 '19 at 5:46
  • 1
    "Do not change the current 4 net work IP addresses." That is, in a word, stupid. This is a business, and if the business depends on its network, then the network should be done correctly. That means it needs to be flexible, and not being able to renumber networks is not flexible, and overlapping networks is incorrect. There is outside-source NAT that should only be used temporarily, until the overlapping networks can be renumbered. This is a situation faced by merging companies, but they do not let it go on for long; they fix the overlapping network addressing. – Ron Maupin Jul 22 '19 at 14:13
  • Also, please quit using network classes. Network classes simply do not exist since 1993 (26 years ago, before the Internet went commercial in 1995!). Modern networking is CIDR (Classless Inter-Domain Routing). – Ron Maupin Jul 22 '19 at 14:54
  • @RonMaupin, sending this across. Please merge with the other post there. – Bhargav Rao Jul 24 '19 at 11:19

You can have multiple solutions

  1. You are not allowed to Change IP addresses, But you can change Subnet mask.
  2. You can use VLANs with Manageable Switch. (Implementation of option 01)
  3. You Can Use VRF. But Generally VRF Use in large Environment.

If you have same IP address range in all division mean you have only one network. That mean you need not router. Switch is enough to connect between all divisions.

If you are not allowed to change even subnet mask, You can enable bridge mode in router and use router as a Switch.

But you should not allow to use same ip address in twice.(It will be caused for IP conflict)

However Best Solution is Readdressing your network.....

I don't understand your third question. I assume that It will solved by mentioned solutions.

How to enable bridge mode

  • 1
    VRF only solves the (potentially) duplicate IP problem on the top level. It doesn't solve the "CEO wants access to everywhere" problem. – Zac67 Jul 22 '19 at 8:18
  • Plus, bridge mode can only work if no addresses collide which the OP can't rule out. – Zac67 Jul 22 '19 at 8:29
  • @Zac67 second point i have addressed. – infra Jul 22 '19 at 8:34
  • Thanks all. Maybe readdressing is the best solution. I will try to do it/. – kissinger Jul 22 '19 at 18:48

The only sane way to solve this problem is renumbering the network. However, in addition to the present answers, you could assign multiple IP addresses to the end nodes. That way, they keep the address you mustn't change but you simply use an additional, unique address for inter-subnet routing. Sadly, this rules out DHCP (unless you use multiple (virtual) NICs).

A variant of renumbering is to go dual-stack and use unambiguous IPv6 subnets for inter-subnet routing. However, migrating to IPv6 is a complex task and probably overkill for this one issue alone.

The alternative for renumbering is to NAT all the way through: you assign each subnet a unique, virtual prefix and DNAT on the last hop from the virtual IP to the local one. If the source subnet (CEO) is colliding with the other addresses as well you also need to SNAT that on the first hop. Overall, this quickly gets very complicated and isn't really a sane way to go.

Yet another alternative is to use proxy ARP to make the remote hosts appear as local ones. For this you need to make sure there are no address collisions. Also, proxy ARP is a pain to work with as well and best avoided.

  • ... assign multiple addresses to an end host, in a company and IT organisation that failed to prevent a network with mutliple adressing overlaps? Technically feasible certainly, but the complexities won't become less (more or less random source address selection by the host for outgoing connections, auto-generated DNS entries pointing to the "wrong" address half of the time...). – Marc 'netztier' Luethi Jul 23 '19 at 6:09
  • @Marc'netztier'Luethi True - if you treat the current addresses as legacy and don't route anything with them (practically link local) and create multiple DNS entries it's feasable. Real renumbering is much more practical, of course - "the only sane way"... – Zac67 Jul 23 '19 at 6:15

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.