schedmd/slurm — reverse-engineered prompt

Build me a Linux cluster workload manager like Slurm. I want admins to describe a group of compute nodes, and users to submit work that needs some of those nodes for a period of time. The system should decide when jobs can run, keep a queue when requests conflict, allocate nodes either exclusively or shared, start jobs on the assigned nodes, and let users monitor what is running or waiting.

Please make it simple to build and install on Linux, with sample config files, command line tools, clear docs, and tests. It should be scalable enough for real clusters, tolerate failures as much as practical, and not assume one specific network setup. Use C for the core system and organize it like a serious open source project with source, headers, docs, config examples, contrib tools, and a test suite.

Look up current Slurm docs online if you need to understand the expected behavior.