Arista EOS AI fabric lab
Practice EOS-inspired AI fabric operations, running versus startup configuration, eight-rail paths, MTU faults and synthetic RoCE congestion monitoring.
7060X6 leaf / 7800R4 spine reference architecture. Finite EOS-inspired commands; no cEOS/vEOS image or switch ASIC is executed.
All switching labs · Model, commands and sources · Cluster lab
Configure the topology, then launch when you are ready to use lab time.
Five minutes to explore. More time when you need it.
Try a hosted lab free for five minutes per verified account. After that, buy an hour for $20 USD. No subscription or automatic renewal.
Use your Wirewalk Institute account. Create an account, confirm your email or reset your password, then return here to sign in. No course purchase is required.
Sign in before launching a lab. Your free time starts only when you launch or resume the lab.
Pause stops time and freezes lab controls. Closing or losing the connection stops time within four seconds. One active lab per account. Progress is saved automatically; viewing it and exporting a session use no time. Access and prepaid-hour details.
Fabric topology and saved sessions
Eight GPU/HCA rails; two leaf roles per rail; two spine roles shared by each plane. Rebuilding replaces only this vendor's saved fabric, not your other labs or purchased time.
Follow the eight rails
Configure a switch
SM service hosts and partition membership
Each SM service host represents one manager instance per independent plane/subnet; these controls affect all its instances together. The displayed master is the common elected host role. Individual per-plane elections are outside this model. 0x9234 = full membership in partition 0x1234; 0x1234 = limited membership. Two limited endpoints cannot communicate. This control represents SM administration, not a host assigning its own membership.
Diagnose, repair, validate
Inject an exercise, inspect a supported diagnostic while the fault exists, repair it, then run a new passing collective check. Repair alone never changes a failed result to passed.
Try one dependency at a time
- Lose one leaf uplink and inspect the remaining path. Lose both uplinks on one rack side and watch that rail become unavailable.
- Interrupt a shared spine, then its partner. Compare the rails affected within the same plane.
- Compare physical link state with logical reachability. A cable being up does not establish MTU, BGP or partition compatibility.
- Use the configuration workflow to repair the cause; inspect the counters and run a fresh check.
Individual faults
Rail selection above applies here. Switch faults use the switch selected in Configuration. Shared spines affect every rail assigned to that plane.
State-aware training console
Finite command subset, synthetic summaries and selected-switch scope. No vendor binary, host shell, SSH or arbitrary command execution. Use the GUI for settings outside the listed subset.
WIREWALK EDUCATIONAL SWITCHING MODEL — synthetic output only.
Monitor the consequences
Queue occupancy, errors, ECN/PFC events and credit wait are teaching counters, not predicted packets, bytes, latency or throughput. Hardware port rates identify a reference topology only. Queue gauges can fall; event counters remain cumulative.