Logging

reconD writes structured logs to stdout. Under the packaged service, they go to the journal:

sudo journalctl -u recond -f
sudo journalctl -u recond --since "1 hour ago"

Format and level

log:
  format: text   # default: JSON
  level: debug   # debug | info | warn | error; default: info

JSON, the default, suits journald and log shippers such as Promtail, Fluentd, or the Datadog agent. text is easier to read during local development.

At info, a quiet workload logs almost nothing. Use debug to see every cycle, including no drift detected and each docker command reconD runs.

Fields

Every line from a workload's reconciler carries workload and container. Slack delivery errors are the exception: they are logged without either. Most lines also carry a phase, which you can filter or alert on:

phase	Meaning
poll	Fetching deployment metadata from GitHub.
drift_check	Comparing desired, recorded, and running digests.
runtime_inspect	Running docker inspect on container_name (logged at debug).
runtime_image_inspect	Running docker image inspect to find the running image's digest (logged at debug).
image_pull	The starting deployment line. reconD does not pull images itself; your deploy command does.
deployment_strategy	Applying the strategy.
restart	Each forward attempt and its deploy_command.
health_check	Waiting for the health check.
commit	Writing state, and prune failures.
image_prune	Running docker image prune after a successful deploy.
rollback	Restoring the previous release, including backoff between rollback attempts.
forward_deployment	Backoff before retrying a forward attempt.
noop	No drift; nothing to do.

Other common fields are digest, git_sha, attempt, max_attempts, drift_reason, duration_ms, and error. When deploy_command fails, its stdout and stderr are included.

Useful queries

With the default JSON format, jq can filter the journal. fromjson? skips any lines that are not JSON:

# Everything one workload did
sudo journalctl -u recond -o cat | jq -R 'fromjson? | select(.workload == "my-app")'

# Only errors
sudo journalctl -u recond -o cat | jq -R 'fromjson? | select(.level == "ERROR")'

# Why reconD decided to deploy
sudo journalctl -u recond -o cat | jq -R 'fromjson? | select(.drift_reason) | {time, workload, drift_reason}'