Hello, I have a Power Server with an IBM i stand-alone system that I'm going to add a client partition to. Both partitions will be IBM i running Windows 7.6 and will have the Operations Console. Sometimes I have problems because if I activate one console, I have to disconnect the other. I only have one Ethernet port on the PC. Why is this happening? What PC do you recommend, or is it necessary to have two Ethernet connections?

Hi Miguel, good to see you back!

First the usual decoding: I assume "Windows 7.6" means IBM i 7.6, and since there is no HMC in your picture I will also assume the client partition will be created with Virtual Partition Manager (VPM), the IBM i hosting IBM i way. Correct me if an HMC is involved, because some details change.

Now, to your "why": the single Ethernet port on your PC is innocent. What you are seeing is a rule of the platform: each partition allows exactly ONE active console device at a time. So if activating one connection forces you to drop the other, those two connections are almost certainly landing on the SAME partition's console. The usual causes are two ACS configurations pointing to the same service host name or IP, or the same serial number plus target partition combination in both configs. Between two DIFFERENT partitions there is no such exclusion: one PC with one NIC can hold multiple console sessions open side by side, and IBM documents exactly that scenario (consoles for multiple systems or partitions from a single PC) in the Operations Console guide:
https://www.ibm.com/docs/en/ssw_ibm_i_71/rzajr/rzajr.pdf

So no, you do not need two Ethernet connections, and honestly any PC that runs a current ACS is fine. The console is just TCP sessions on ports 2300, 2323, 3001 and 3002 towards different service IPs (ports reference: https://www.ibm.com/support/pages/how-check-firewall-blockage-operations-console-or-hmc-5250-console). This is a configuration story, not a hardware one.

Here is the setup that makes host and client consoles coexist peacefully:

1.⁠ ⁠Each partition gets its OWN service tools LAN adapter, with its OWN unique service IP and service host name. Two partitions sharing a service IP is exactly the recipe for the mutual kick-out you described.

2.⁠ ⁠The host partition uses a physical port for its console, as today. The client partition has no physical hardware, so its console rides a VIRTUAL Ethernet adapter, bridged to the real network with Ethernet layer-2 bridging on the host: two line descriptions with the same BRIDGE identifier, one on the physical resource and one on the virtual one. The official IBM page covers the whole procedure:
https://www.ibm.com/support/pages/creating-ibm-i-or-linux-client-partitions-using-virtual-partition-manager-vpm
and the classic Redpaper walks it step by step:
https://redbooks.ibm.com/abstracts/redp4806.html

3.⁠ ⁠And here is the pitfall buried in that official page, which fits your symptom suspiciously well: the physical resource you pick for bridging must NOT be in use by any varied-on line description, by the LAN console, or by remote support. In other words, do not bridge the same physical port that carries the host's own LAN console. On small boxes with few ports this is where the planning matters: give the host console its port, and bridge a different one for the client traffic and the client console.

4.⁠ ⁠In ACS, create two separate console configurations, each with its own service host name, its own IP, and the correct serial plus target partition. From that moment both console windows can stay open at the same time from your single NIC.

If you share whether it is really VPM (no HMC), how many physical Ethernet ports the server has, and the two service IPs you plan to use, I can help you sanity check the port layout before you build the client partition. That is much cheaper than debugging it afterwards.

Cheers
Roberto