ihc doctor

Is the cloud on the air, and if not, what to do. One command that checks the box the way the smoke tests do, read only, in a few seconds, and says for each problem what to run next.

cd /opt/inhousecloud && ./ihc doctor           # from the install directory
./install.sh doctor                            # from the checkout, the same
scripts/ihc-doctor.sh --install-dir /opt/inhousecloud --json   # for an agent or a script

--verbose adds the evidence of each check; --json prints {checks, summary, generated_at} for an agent (operator-contract.md). The exit code is 1 when a check fails.

What it checks

The install and state.json, Docker, the containers, disk and memory, the proxy's routes, DNS and TLS for the ingress, the sign-in, Gitea and its runners, Postgres, Redis, the backup's age, every app in the inventory (its health path through the proxy), the operator agent when the agent profile is on, and how far this checkout is from origin/main. Each check answers ok, warn or fail in one line, and for warn and fail the action and the runbook that goes with it (runbooks).

When the portal answers, ihc doctor --json and install.sh status --json read the platform's health from the portal API and say "source": "portal-api"; otherwise "local".

Which installation it checks: --install-dir, else $IHC_INSTALL_DIR, else the current directory when it is one, else /opt/inhousecloud, else the directory the running Compose project was started from.

Inside the operator agent

The operator agent (the agent profile) has no Docker socket and no shell on the host. It runs ihc-ro doctor inside its container: the checks it can make through the read-only Docker API and the routes, plus the host's last report.

How it was tested

On a clean VM, then with three failures caused on purpose (an app's container stopped, its proxy route deleted, the backup sidecar stopped and its last success made three days old): each was named in one line with the right action, and fixed by that action. About 3 seconds a run. The record: doctor run.