sonic-kvm-testbed_skill

Overview

This skill deploys and manages a local SONiC sonic-mgmt KVM virtual testbed with converged cEOS neighbors and a sonic-vs DUT. It automates provisioning of T0, T1, and T1-LAG topologies on a single host for pytest-based network validation. Use it to create, verify, run tests against, troubleshoot, and tear down a repeatable virtual SONiC lab.

How this skill works

The skill orchestrates a KVM VM for the sonic-vs DUT and Docker containers for cEOS neighbors, PTF traffic generator, and the sonic-mgmt framework. It configures a management bridge (br1), imports images, launches containers/VMs with testbed-cli.sh, deploys minigraph, and verifies BGP and container status. It includes fix scripts to reapply critical inventory and credential changes that revert during operations.

When to use it

• Set up a local KVM-based SONiC testbed for functional or integration pytest runs
• Deploy T0, T1, or T1-LAG topologies with converged cEOS neighbors on one host
• Troubleshoot testbed-cli.sh add-topo or deploy-mg failures and configuration reverts
• Tear down and redeploy testbeds repeatedly during development or CI debugging
• Run sonic-mgmt test cases against a virtual SONiC switch (sonic-vs) without physical hardware

Best practices

• Run the provided fix-configs script before every testbed operation to restore critical settings
• Use docker import for cEOS images (do not use docker load) and store images locally
• Set max_fp_num to 127 in vm_host vars to avoid port/FP limits on larger topologies
• Remove vars/topo_<TOPO>.yml.bak before add-topo to prevent KeyError in the converger
• Enable use_converged_peers in vtestbed.yaml only with the required sonic-mgmt master PR
• Persist the br1 management bridge across reboots via netplan or system network config

Example use cases

• Deploy vms-kvm-t0 topology to validate BGP and ACL behavior with a single cEOS converged peer
• Launch vms-kvm-t1-lag for LAG and partial spine convergence testing with 24 raw neighbors
• Reproduce and debug deploy-mg failures locally by inspecting DUT boot user and sudo setup
• Run pytest suites from sonic-mgmt container against sonic-vs DUT for feature validation
• Tear down and rebuild a testbed to test automation scripts or CI gating changes