Troubleshooting
Why Two Bridged VMs Cannot Ping Each Other in a PLC Simulation
Short answer
Both bridged guests carried the same MAC address, so the bridge could not forward frames between them: VM-to-VM and VM-to-simulated-PLC pings failed while every host-side test passed. Generating a new MAC address on one guest restored communication; distinct IPs, a shared subnet mask and ICMP-friendly firewall settings are the follow-up checks.
On this page
- Why can VM #2 ping the host but not VM #1?
- What does the ping pattern prove about the fault?
- How do two bridged VMs get the same MAC address?
- Why does one shared MAC stop guest-to-guest traffic?
- How do you give each VM its own MAC address?
- What else must you verify after the MACs differ?
- How do you download the HMI project to the runtime VM?
- Is two runtimes in two VMs a sound simulation layout?
- Questions asked
Both guests were bridged to the host's physical NIC, and both carried the same MAC address — that is why VM #2 could not ping VM #1 or the simulated PLC behind it. The host at 10.1.0.250, the simulation VM at 10.1.0.251, the HMI runtime VM at 10.1.0.252 and the simulated PLC at 10.1.0.1 passed every host-side test, because none of those tests crossed the bridge between the two guests. Generating a new MAC address on one guest restored both failing paths. On a cloned pair, compare the two adapter MACs before you touch anything else: it is the most likely cause and the cheapest to confirm.
Why can VM #2 ping the host but not VM #1? #
Because both guests carried the same MAC address: VM #2 at 10.1.0.252 could reach the host at 10.1.0.250, whose interface has its own address, but every frame meant for VM #1 at 10.1.0.251 ran into a duplicated identity on the same bridge.
The layout itself is valid. Three systems shared one subnet — the host running TIA Portal, VM #1 running S7-PLCSIM Advanced with its simulated controller at 10.1.0.1, and VM #2 running the WinCC Unified runtime — with both adapters set to Bridged Adapter on the host's physical NIC. A bridged guest behaves like any other device plugged into that network, and two bridged guests can talk to each other through the switch.
The clue is in which tests passed. The host pinged the simulated PLC and downloaded to VM #1; VM #2 pinged the host. Every one of those paths ends at an interface with a unique MAC. The two failing paths — VM #2 to VM #1, and VM #2 to the simulated PLC — both required forwarding toward the duplicated identity. That reading points at the guests' adapter identities, not at the host configuration.
What does the ping pattern prove about the fault? #
The pattern isolates the fault to guest-to-guest forwarding across the bridge: every test that involved only the host's own interface succeeded, and every test that had to pass between the two guests failed.
| Test | Result | What it proves |
|---|---|---|
| Host 10.1.0.250 to simulated PLC 10.1.0.1 | Replies | The host's path into the simulation works. |
| Host download to VM #1 at 10.1.0.251 | Succeeds | The host reaches the simulation VM through the bridge. |
| VM #2 at 10.1.0.252 to host 10.1.0.250 | Replies | VM #2's bridged adapter works. |
| VM #2 to VM #1 at 10.1.0.251 | No reply | Guest-to-guest forwarding hits the duplicate. |
| VM #2 to simulated PLC 10.1.0.1 | No reply | The same broken path blocks the HMI-to-PLC traffic. |
The decisive observation came from the hypervisor, not the network: both guests showed the same MAC address in their adapter settings. A new address was generated for one guest and both failing tests began to reply — that is the confirmed cause here. A failed ping can also come from a subnet-mask mismatch, a duplicate IP, a disconnected virtual adapter or a firewall rule dropping ICMP; those are the follow-up checks once the MACs are known to differ.
How do two bridged VMs get the same MAC address? #
Cloning does it: a copied guest carries the copied adapter settings with it, so unless the clone step regenerates the address, both guests present one identity to the physical network.
VirtualBox's clone wizard offers an option labeled "Reinitialize the MAC address of all network cards"; leave it unticked, or copy a VM folder by hand instead of cloning, and the copy keeps the source adapter's address. Copying duplicates more than the MAC: a cloned Linux guest can reuse the source machine ID, which the DHCP server reads as one client identifier, so two guests can draw the same lease. Neither duplicate announces itself, and most traffic keeps working — which is exactly what makes this fault hard to see.
Why does one shared MAC stop guest-to-guest traffic? #
Because Ethernet bridges forward on MAC addresses: IEEE 802.1D address learning records the port a source address arrives on, and one address shared by two guests makes that learned port flap between them, so frames reach the wrong guest or neither.
An IP address identifies a host at the Internet Protocol layer; a MAC address identifies an interface at the Ethernet layer. Before one system sends an IP packet to another on the same subnet, address resolution maps the destination IP to a MAC, and the frame is forwarded using that MAC. Learning assumes one address lives on one port. Two guests sourcing frames from the same MAC on different ports break that assumption, and resolution can bind two IPs to one link-layer identity — frames meant for 10.1.0.251 then land at the guest configured as 10.1.0.252, or nowhere. The partial functionality is the signature: paths to unique interfaces worked, paths to the shared one did not. The same signature appears on other hypervisors, as covered in troubleshooting a VMware bridged guest that cannot ping.
How do you give each VM its own MAC address? #
Power off both guests, open each one's network settings, expand the Advanced section, and use the control that generates a new MAC address on one guest so the two values differ — then boot both and retest.
- Shut down both guests; adapter identity applies cleanly only while a guest is powered off.
- Open VM #1's network settings and confirm the adapter is a bridged adapter on the host's physical NIC.
- Expand Advanced and read the MAC address shown for the adapter.
- Repeat for VM #2 and compare. If the values match, generate a new MAC address for one guest and save.
- Start both guests and confirm inside Windows that the physical addresses differ —
ipconfig /alllists the address under each adapter. - From VM #2, ping 10.1.0.251; once it replies, ping 10.1.0.1. Both should answer.
- Next time you clone, tick "Reinitialize the MAC address of all network cards" in the wizard, or compare the two MACs before the copy's first boot.
Change one guest only. One new identity is enough to break the collision, and it leaves half the installation known-good.
What else must you verify after the MACs differ? #
Three things, in order: distinct IP addresses, one subnet mask on the host and both guests, and a firewall profile that allows ICMP echo — a unique MAC fixes none of these.
Distinct IP addresses. A unique MAC does not prevent an IP conflict. Confirm VM #1 holds 10.1.0.251 and VM #2 holds 10.1.0.252, statically or by reservation, because a clone reusing a machine ID can draw a duplicate DHCP lease under a different MAC. On Linux guests, regenerating the machine ID is the documented fix for the duplicate client identifier.
One subnet mask. Verify the mask on all three systems instead of inferring it from the addresses. With 255.255.255.0, the network of each system is its address ANDed with the mask:
- 10.1.0.250 AND 255.255.255.0 = 10.1.0.0 (host)
- 10.1.0.251 AND 255.255.255.0 = 10.1.0.0 (VM #1)
- 10.1.0.252 AND 255.255.255.0 = 10.1.0.0 (VM #2)
- 10.1.0.1 AND 255.255.255.0 = 10.1.0.0 (simulated PLC)
All four sit on 10.1.0.0/24, so no ping between them needs a gateway. Change one system's mask to 255.255.255.128 and that stops being true: 10.1.0.250 AND 255.255.255.128 = 10.1.0.128, while 10.1.0.1 AND 255.255.255.128 = 10.1.0.0 — different networks, failed pings, nothing wrong at the bridge.
Firewall and binding. If pings fail while a download or a runtime connection succeeds, suspect the guest's firewall dropping ICMP echo, not the bridge. And after any adapter change, reconfirm that the PLC simulation is still running in TCP/IP mode bound to the "Ethernet" adapter it was configured for.
One caution that belongs here: a bridged guest is a real device on whatever network the host's NIC belongs to. Run this bench on an isolated lab switch or VLAN — a duplicate MAC or IP introduced on a plant network disturbs live controllers, not just the simulation.
How do you download the HMI project to the runtime VM? #
Confirm the host can reach 10.1.0.252, then start the download from TIA Portal, let the extended download dialog open, and select the runtime VM by name or IP address rather than Profinet discovery.
With the MACs unique, treat the runtime VM the way you would treat a panel:
- From the host, ping the runtime VM to confirm the path.
- Start the transfer in TIA Portal and open the extended download dialog.
- Select the target by name or IP address; a bridged VM's discoverability depends on the physical network, and direct addressing does not.
- Where the dialog accepts it, enter the VM's address, 10.1.0.252, as the target.
- Review the target line, transfer, and confirm the project starts in the guest's runtime.
If the dialog finds nothing while ping works, the bridge is not the problem: confirm the runtime is installed and running in the guest, that host and guest firewalls allow the download traffic, and that the project's device version matches the installed runtime. For the wider question of host options, see where WinCC Unified Runtime can run.
Is two runtimes in two VMs a sound simulation layout? #
Yes — for a simulation bench, one runtime per guest is a workable way to run two panels at once, and it is the right call when a single host cannot run both runtimes as the project needs; the conditions are identity, licensing and memory, not networking.
Each guest needs its own runtime installation and its own license, the project's device configuration has to match the installed runtime version, and the host must carry three operating systems at once — TIA Portal on the host plus two Windows guests — so size memory for that before blaming the network for sluggish screens. Network reachability proves none of this; it only proves the path. Keep the bench on the isolated subnet and record the TIA Portal, runtime, Windows and hypervisor versions in use, because a later download or runtime fault gets compared against exactly that list.
Keep reading: testing HMI simulation against a live PLC and diagnosing intermittent HMI to PLC Ethernet comms hangs.
Questions asked
Does generating a new MAC address change anything else in the guest?
Windows sees the adapter as a new network connection, so expect the network location prompt again and re-check which firewall profile applies. The adapter's name inside the guest does not change, so software bound to that name keeps working. If any license installed in the guest keys on the MAC address, plan for that reactivation before you regenerate.
Can both simulated panels run on one host PC instead of two VMs?
Only if the installed runtime lets you run two instances with the settings the project needs, which is exactly what this layout works around. Two VMs sidestep the single-instance limit at the cost of running three operating systems: budget memory for the host with TIA Portal plus two Windows guests, and give each runtime its own license.
Would NAT adapters avoid the duplicate MAC problem?
They hide it rather than fix it. Under NAT the guests stop appearing on the 10.1.0.x subnet the simulated PLC lives on, so host downloads and guest-to-guest IP traffic behave differently and direct IP targeting stops working. Bridged adapters with unique MACs keep every system on one subnet, which is what this simulation needs.
How do you confirm the two VMs really have different MAC addresses?
Compare the MAC fields in both guests' network settings in the hypervisor while they are powered off, then check inside each running guest with ipconfig /all and compare the physical addresses. Doing both catches the case where a regenerated address was saved in the hypervisor but the guest still runs the old value.