agentusecasesAll 965 use cases
Software & DevOps

Diagnose a killed process on a server

Codex CLI over SSH found the OOM killer in kernel logs on a 2 GB droplet with no swap and listed fixes by risk without changing anything.

Done withCodexClaude Code

What they did
Claude Code kept dying within seconds on the author's Ubuntu droplet. He ran Codex CLI on his Mac and gave it the SSH command for the server, plus the symptom and no hints. Codex asked approval to connect, then checked memory, processes and limits. It grepped the kernel log via journalctl for kill and OOM messages.
How it went
Codex found the OOM killer ending claude at about 860 MB on a 1.9 GB droplet with no swap, with MySQL using about 786 MB. It ranked fixes and changed nothing.
Worth knowing
The author resized the droplet, which caused its own problems. He says in hindsight that the first fix Codex listed, a 2 GB swap file, was cheaper.

Try it yourself with Claude Code

My server at [user@host] has a process that keeps getting killed, namely [service name]. Connect over SSH and investigate read-only: check kernel logs, memory, swap, and service logs to find the cause. Finish with a diagnosis and a list of possible fixes ranked by risk. Do not change any settings, install anything, or restart services without showing me the exact command and getting my approval first.

Read the original ↗

Source: flaviocopes.com · Undated

Five of these in your inbox every morning

The best things people got an AI agent to do, each with the prompt to try it.

More like this