In this Field Note
Stop the router in the middle. If you can predict which path the packets take next, explain why, and prove the application recovered, you have learned more than a roomful of green status lights can tell you.
This GNS3 exercise uses three MikroTik CHR routers and two VPCS endpoints. It deliberately prefers a two-link route through R2 over the apparently shorter direct link. Then you break that preferred route and investigate the result.
It is a learning configuration for isolated, fresh lab devices—not a production configuration or a throughput benchmark. The commands follow the RouterOS 7 documentation; they have not been executed on your chosen CHR build. Record and verify that build as part of the exercise. The failure-rehearsal article and downloadable evidence lab provide the method for recording what happens.
Set up a lab that cannot surprise the office
Use compatible GNS3 components and a QEMU-capable x86-64 server or GNS3 VM. An ARM laptop may need a compatible remote x86-64 lab host; do not assume an x86 disk image will run efficiently through your local hypervisor. Check the official CHR appliance definition.
Obtain the image from MikroTik and the appliance template from GNS3. Record the RouterOS version and image checksum, use the same version on all three routers, and confirm licensing. MikroTik’s free CHR license limits upload to 1 Mbps per interface. That is enough for small routing probes. It is emphatically not a referendum on the speed of MikroTik hardware.
Create fresh R1, R2, and R3 instances and VPCS nodes PC-A and PC-B. Set unique administrator passwords. Do not add a Cloud, NAT, or physical bridge connection: this exercise needs no internet connection inside the topology. Keep console access available. Do not paste these commands into an existing router or clear a production configuration to make it resemble the lab.
Wire it before configuring it
| Link | First end | Other end |
|---|---|---|
| R1–R2 | R1 ether1: 10.0.12.1/30 | R2 ether1: 10.0.12.2/30 |
| R2–R3 | R2 ether2: 10.0.23.1/30 | R3 ether1: 10.0.23.2/30 |
| R1–R3 | R1 ether2: 10.0.13.1/30 | R3 ether2: 10.0.13.2/30 |
| LAN A | R1 ether3: 192.0.2.1/24 | PC-A: 192.0.2.10/24 |
| LAN B | R3 ether3: 198.51.100.1/24 | PC-B: 198.51.100.10/24 |
The LANs use documentation address space; transit and loopback addresses use private space. Give R1, R2, and R3 loopbacks 10.255.0.1/32, 10.255.0.2/32, and 10.255.0.3/32 respectively. An empty bridge named lo supplies the loopback interface here.
The preferred transit interfaces have OSPF cost 10. The direct R1–R3 interfaces have cost 100. Before touching the console, write down which route should win from each end and what should replace it if R2 disappears.
Configure one router at a time
Check interface names with /interface print. Apply each block once to its matching fresh router. Repeating it can create duplicate objects or errors. The RouterOS OSPF interface-template reference is the syntax authority; stop and check your version if a command is rejected.
R1
/system identity set name=R1
/interface bridge add name=lo
/ip address
add address=10.255.0.1/32 interface=lo
add address=10.0.12.1/30 interface=ether1
add address=10.0.13.1/30 interface=ether2
add address=192.0.2.1/24 interface=ether3
/routing ospf instance add name=lab version=2 router-id=10.255.0.1
/routing ospf area add name=backbone instance=lab area-id=0.0.0.0
/routing ospf interface-template
add area=backbone interfaces=ether1 type=ptp cost=10
add area=backbone interfaces=ether2 type=ptp cost=100
add area=backbone interfaces=lo passive=yes
add area=backbone interfaces=ether3 passive=yes
R2
/system identity set name=R2
/interface bridge add name=lo
/ip address
add address=10.255.0.2/32 interface=lo
add address=10.0.12.2/30 interface=ether1
add address=10.0.23.1/30 interface=ether2
/routing ospf instance add name=lab version=2 router-id=10.255.0.2
/routing ospf area add name=backbone instance=lab area-id=0.0.0.0
/routing ospf interface-template
add area=backbone interfaces=ether1 type=ptp cost=10
add area=backbone interfaces=ether2 type=ptp cost=10
add area=backbone interfaces=lo passive=yes
R3
/system identity set name=R3
/interface bridge add name=lo
/ip address
add address=10.255.0.3/32 interface=lo
add address=10.0.23.2/30 interface=ether1
add address=10.0.13.2/30 interface=ether2
add address=198.51.100.1/24 interface=ether3
/routing ospf instance add name=lab version=2 router-id=10.255.0.3
/routing ospf area add name=backbone instance=lab area-id=0.0.0.0
/routing ospf interface-template
add area=backbone interfaces=ether1 type=ptp cost=10
add area=backbone interfaces=ether2 type=ptp cost=100
add area=backbone interfaces=lo passive=yes
add area=backbone interfaces=ether3 passive=yes
On PC-A, enter ip 192.0.2.10/24 192.0.2.1 and save. On PC-B, use ip 198.51.100.10/24 198.51.100.1 and save.
Prove the boring state first
Run /routing ospf neighbor print and inspect /ip route print. Each router should have two Full neighbors when all transit links are working. From PC-A, ping 198.51.100.10 and run trace 198.51.100.10; test the reverse direction from PC-B too.
R1 should choose 10.0.12.2 as the next hop toward LAN B. Its transit cost through R2 is 10 + 10, versus 100 on the direct link; the destination LAN cost is common to the comparison. OSPF is minimizing the configured cost, not the number of lines on your drawing.
If a neighbor is missing, inspect cable endpoints, interface state, subnet masks, OSPF area and template matching before changing timers. If neighbors are Full but the PCs cannot communicate, inspect learned LAN routes and PC gateways. A default route added to hide a missing internal route would make the experiment less useful.
Save a clean project and configuration exports once the baseline works.
Break one thing, and keep the receipts
- Run repeated probes from PC-A and record the initial route and trace.
- From R2’s lab console, run
/interface disable ether2. Record the action time. - Observe neighbor state, route selection, lost probes, and the new trace. R1 should switch toward 10.0.13.2.
- Run
/interface enable ether2on R2. Confirm the preferred route and both-direction connectivity return. - Restore the baseline. Stop R2 entirely and repeat. Compare the observed detection and recovery sequence rather than assuming the two faults behave identically.
Record the first observed successful probe separately from the route-change event. Your probe interval limits timing precision; a successful ping also does not prove an application transaction or a large packet works. The failure-evidence lab makes those distinctions explicit.
Give the result a useful boundary
Change a cost and predict the route before testing it. Next, investigate a silent packet drop with the physical interface still up. If you add BFD, follow the vendor’s session and OSPF configuration guidance, verify that a session is actually established, and record negotiated behavior. A checked option is not a timing measurement.
This exercise can expose addressing, route-preference, and recovery mistakes. It cannot establish RF performance, optical budgets, PoE compatibility, hardware forwarding behavior, or production convergence time. Leave that distinction in the test notes.
For a commercial migration, the next step is to replace the teaching question with the real one: which service must survive, which policies must hold, and how will the operator recover control? Oaktown Tech can scope a focused model and repeatable test procedure. Three routers are enough to learn the discipline. The number of icons is not the deliverable.

Join the discussion