A useful virtual lab must produce evidence, not just a working topology
GNS3 and Wireshark solve different parts of a networking exercise. GNS3 provides a workspace where supported simulated or emulated devices can be connected and tested. Wireshark decodes traffic captured at an observation point. A successful lab uses both: configure a deliberately small network, predict the packets that should appear, record the actual packets, introduce one controlled fault, and explain the difference. Building a large virtual topology without a question to answer usually teaches less than a two-device exercise with clear evidence.
GNS3 is not itself a source of unrestricted Cisco operating-system images. Appliance emulation may require appropriately licensed images, enough host resources, compatible hypervisor settings, and in some cases a GNS3 VM. The built-in VPCS endpoints and Ethernet switch are a sensible starting point because they do not require a separately acquired network OS. The details below are a reproducible lab procedure, not a claim that a particular topology was launched or tested on this site. Use only equipment, images and network access for which you have permission.
Before creating a lab, write down three expected observations. For an IPv4 same-subnet ping, the sender may first resolve the destination’s MAC address using ARP, then exchange ICMP Echo messages if the path and endpoint allow them. A second ping may not produce another ARP request if the mapping is already cached. If the destination is moved into a different subnet without a router or suitable default gateway, the forwarding decision changes. These predictions give the packet capture an instructional purpose instead of turning it into a screen full of unrelated frames.
Build the smallest topology and label its assumptions
Open a new GNS3 project and place one built-in Ethernet switch and two VPCS endpoints, named PC1 and PC2. Connect each VPCS device to the switch, then start the nodes and open their consoles. The example uses isolated private lab addresses: on PC1 enter ip 10.10.10.11 255.255.255.0, and on PC2 enter ip 10.10.10.12 255.255.255.0. The exact command syntax shown here is for VPCS; it is not intended for Linux, Windows or Cisco IOS shells. Depending on the GNS3 and VPCS versions, consult the node help for supported commands before troubleshooting an apparent syntax error.
Use the VPCS console to verify the configured address and mask, and use save to persist each endpoint configuration, then initiate ping 10.10.10.12 from PC1. If both nodes are running and the links are up, that is the expected connectivity test, but a missing reply is a diagnostic input, not proof that GNS3 is broken. Recheck the project state, the endpoint addressing, cable/link connections and local node status before adding more devices. Do not introduce a router, DHCP server or Internet connection merely to make the topology more impressive.
Place a short note in the project documenting the intended model: one Ethernet broadcast domain, two directly connected IPv4 hosts, no gateway, and no remote network. This is important because learners often over-interpret an isolated switch topology as proof about real access-switch VLAN enforcement, spanning-tree reconvergence, enterprise Wi-Fi or a routed WAN. The built-in switch is a teaching aid; it cannot validate every capability of an actual managed switch platform.
Capture on the correct link and preserve the raw evidence
GNS3’s supported interfaces allow packet capture from a link—typically by right-clicking that connection and choosing Start capture. The packet viewer workflow depends on the installed GNS3 release and local or web client integration. If Wireshark does not launch automatically, first determine where the capture file is saved and whether the local console and capture protocol handlers are configured. A missing application window does not mean no traffic was sent. Capture on the PC1-to-switch link for a clear view of PC1’s transmitted and received frames.
Begin the capture before the first ping if you want to see initial neighbor resolution. In Wireshark, use display filter arp || icmp to focus on the relevant exchange. This is a Wireshark display expression, not a libpcap capture filter. The ARP request is typically a broadcast asking which device has 10.10.10.12; the response supplies the MAC address. ICMP Echo Request and Echo Reply then show an IP-level reachability test. Inspect Ethernet source/destination addresses separately from the IP source/destination addresses: the two layers answer different questions.
Save a copy of the raw packet capture and record the test order and time window. If the capture begins late, a cached ARP entry may explain why ARP is absent. You can make a second observation after clearing neighbor state or restarting a node, if the supported node commands permit it. Do not fabricate an ARP capture by adding annotations that imply a packet existed when it was not seen. The packet-level Wireshark workflow is useful for examining the relationship between frame structure, protocol fields and troubleshooting conclusions.
Introduce a controlled fault and predict the resulting frames
First, deliberately misconfigure the destination address: change PC2 from 10.10.10.12/24 to 10.10.20.12/24, leaving PC1 unchanged and providing no default gateway. Ping the old address again. PC1 may still attempt ARP for 10.10.10.12 because that destination falls inside its own configured subnet, but PC2 no longer owns that address. Observe whether an ARP response is absent, whether an old cached entry persists temporarily, or whether the test behaves differently after neighbor-state expiration. The reason for the missing reply comes from comparing the configured endpoint state with the captured traffic, not from treating every timeout as evidence of packet loss in transit.
Next, restore PC2’s original IP address and verify connectivity. Then try to ping 10.10.20.12 from PC1 without a gateway; the sender must determine that the destination is not on-link. Depending on the VPCS implementation, the console may report a missing route or another local failure rather than emitting the same ARP behavior. This is a controlled demonstration of why a subnet mask and next-hop decision matter. It should not be described as a complete router study, since there is still no real Layer-3 forwarding device in the topology.
To extend the lab, add a licensed and supported routing appliance, place each VPCS endpoint on a different subnet, and configure the appropriate interface addresses and default gateways. Write down the expected ARP and next-hop MAC addresses on each segment before capturing. A router generally re-encapsulates frames for each Layer-2 hop while the end-to-end IP addresses remain the same, absent NAT. Capture separately on both sides; one capture point cannot prove what happened on a different link. For more context, inter-VLAN routing develops the difference between local switching and Layer-3 forwarding.
Extend into DNS and TCP without confusing visibility with diagnosis
A more advanced lab can add a compatible Linux appliance or server that supports DNS and TCP-based services. Keep it inside the isolated virtual topology. Generate one DNS query, then compare the time of the query and response, the requested name, the response status and the resulting address. Use Wireshark’s dns display filter for the initial observation. A positive DNS answer does not guarantee the destination application works, and a name-resolution timeout should be distinguished from a successful negative DNS response. If the lab uses a stub resolver, cache or proxy, record where the capture was taken so you do not mistake local resolver traffic for a query sent all the way to an authoritative server.
For TCP, start a permitted service on one lab node and connect from another. Filter with tcp or a narrower tcp.port == <port> expression in Wireshark, replacing the placeholder with the actual port number. Observe the SYN, SYN-ACK and ACK sequence when a new connection opens, then identify application data and orderly or abrupt teardown if present. Retransmission indicators can be useful, but Wireshark’s expert annotations are interpretations of the captured stream and are sensitive to observation position, missing packets and offload behavior. A TCP retransmission label by itself is not proof that a network device dropped a packet.
When the lab includes both IPv4 and IPv6, treat them as separate protocol exercises rather than assuming a single ping verifies dual-stack readiness. IPv6 neighbor discovery uses ICMPv6 rather than ARP, routers advertise information differently from IPv4 DHCP-based configurations, and link-local addressing requires attention to scope and interface. Record whether IPv6 routing is configured before interpreting an unreachable off-link address. Protocol-specific examples give clearer learning outcomes than enabling every virtual service simultaneously.
Keep the troubleshooting record falsifiable
For each run, preserve the topology assumptions, source and destination addresses, protocol observations and one sentence stating whether they matched the prediction. A simple evidence table can track test, expected result, observation, likely explanation and remaining uncertainty. For example: "After PC2 moved to a different subnet, a ping to the old on-link address did not complete; PC2 no longer owned the destination address." That is defensible if configuration screenshots and the capture support it. "The router failed" is not defensible in a lab that contains no router.
Use a hypothesis sequence: verify power and link, inspect addresses and masks, determine local versus remote destination, inspect neighbor resolution, check forwarding, then test the relevant service. Save original captures before filtering or exporting excerpts. Repeat a failed step after changing only one variable so you know which change affected the result. This is consistent with network troubleshooting by failure sequence, where evidence from each layer constrains the next question rather than creating a long generic checklist.
There are also limits on what a virtual lab can validate. An emulated device may not reproduce hardware forwarding ASICs, real RF propagation, physical optics, precise line-rate behavior, or production control-plane scale. Throughput results depend on host CPU, hypervisor, available images and topology design. The honest outcome of this workflow is a documented protocol demonstration—not a benchmark or a replacement for site testing.
Make a meaningful progression plan
Begin with two same-subnet endpoints and one capture demonstrating ARP plus ICMP. Add one addressing fault and verify the different result. Progress to two IP networks with a router and capture the changed Ethernet next-hop addresses; next introduce a resolver and simple TCP service, capturing each at a known link. Only after these fundamentals are repeatable should you add dynamic routing, VLAN boundaries or dual-stack failure cases. Record one concrete observation for each stage.
The value of GNS3 paired with Wireshark is not that a lab looks like an enterprise diagram. It is that you can ask a precise question, reproduce an event, inspect the evidence and revise your model. That habit transfers beyond certification study into operating real networks, where incomplete captures, cached state, hidden dependencies and incorrect assumptions often cause more confusion than the underlying protocol itself.