For NetOps and Automation Engineers

Details

Where does the first playbook run when the network doesn’t exist yet?

Over the console server. A new rack has no routed production network on day zero, but the SLC 9000 is already on its own uplinks and registered to Percepxion over an outbound-only MQTT connection. Every console in the rack is reachable before anything behind it has an IP, which means automation can build the network instead of riding on top of it.

Smart Groups do the first pass. A new console server calls home, Percepxion matches it to a group, and that group’s firmware and configuration land without anyone touching it. Your playbooks pick up from a known baseline.

What’s actually in the toolchain?

An Ansible collection, two REST APIs and two MCP servers. That’s the whole list, and each piece talks to an API rather than a terminal.

• lantronix.oob 1.2.0. Certified and published on Red Hat Automation Hub, with 22 modules. Some work the Percepxion control plane for fleet operations, others manage a single SLC 9000 directly.
• SLC 9000 REST API, version 2.0. Configuration and operations on one unit.
• Percepxion REST API. Fleet state, jobs, and a fully queryable event log you can pull into whatever your team already runs.
• MCP servers. Percepxion’s has 38 tools and works at fleet scale. The SLC MCP server has 33 tools and works against one unit, which makes it the right fit for lab work and config development.

Plain SSH still works for scripts that need it. Console port N answers on TCP 3000+N.

How do you keep routine chores out of the pipeline?

Hand them to the appliance and the platform. For Cisco IOS and IOS-XE devices defined as managed devices, the SLC 9000 runs config pulls and pushes, health checks and interface actions on a schedule, on the console server itself. Console access stays vendor-neutral for the rest of the rack.

Percepxion covers the fleet side. It schedules jobs across every console server, manages firmware, applies config templates, and backs up and restores appliance configuration. Your pipeline calls those jobs. It doesn’t reimplement them.

For anything that has to live next to the gear, the SLC 9000 runs containers in a sandbox capped at 256 MB of memory and one CPU core. There’s no shell on the firmware, so a container is the supported place for a collector or an agent, and it can’t starve the console ports.

The end state is simple. Out-of-band stops being a parallel manual system and becomes one more target in the same repo, reviewed and versioned like everything else you ship.