Canonfig
Reference

Diagnostics

Run bounded health probes and map failures to the correct recovery path.

doctor runs seven isolated, bounded probes. One timeout or failure does not prevent the other probes from reporting.

canonfig doctor --no-input --timeout-ms 5000

Use JSON when another tool will inspect the report:

canonfig doctor --no-input --timeout-ms 5000 --json

The report schema is canonfig.doctor/v1. Its overall status is healthy, degraded, or unhealthy.

Probes

ProbeWhat it checks
runtimeNode version, Canonfig version, platform, and architecture
stateSQLite header, current migrations, readability, and writability
credentialsSecure noninteractive, explicit local-file, or unavailable capability
sourceConfiguration, reachability, TLS pin, and follower authentication
schedulerNative job state and rendered mechanism
package-managersAvailability of npm, pnpm, or yarn
agent-adapterPolicy, supported adapter, and executable availability

warning and skipped produce a degraded report. fail produces an unhealthy report and maps to a CLI failure category.

Diagnose by symptom

Authentication or revocation

Check whether the Source Machine revoked the Follower Identity, whether the local credential reference still resolves, and whether the stored TLS pin matches the source. Do not reset trust automatically.

Transport

Confirm the loopback source process is running on the enrolled endpoint and that the timeout is long enough for the probe. A reachable endpoint must still pass TLS pinning and authentication.

Scheduler

Inspect native scheduler state:

canonfig schedule status

A state of drifted or disabled is visible; setting the intended calendar can update the native definition.

Agent adapter

Inspect local policy and harness separately:

canonfig agent policy
canonfig agent harness

deterministic-only needs no adapter. Agent proposal or apply needs a supported, available executable.

Synchronization state

canonfig status
canonfig sync --plan

Status explains enrolled identity and recorded run state. A fresh plan rechecks Observed State; it does not erase previous action evidence.

Preserve evidence

Do not delete the SQLite database, cache, or action journal as a first diagnostic step. Those records distinguish interrupted work, drift, and verification failure.

On this page