#Logs

Top-level logs: block. Controls how snapr writes log files to disk.

snapr always emits two streams in parallel:

• stdout — human-readable lines with ANSI colour, never disabled, never written to disk by snapr itself
• files — JSON-line events on disk, rotated by snapr; this is what the API and UI read from

Web UI panels (System log, per-job log) read from these files. Disabling a file disables the corresponding UI panel — snapr returns 503 from the matching API endpoint and the UI shows a "log is disabled" placeholder.

#Layout on disk

When both flags are on (the default), the layout is:

<logs.path>/
  snapr.log                 # all events
  jobs/
    <job-name>.log          # events tagged with job=<name>

Job-name characters outside [A-Za-z0-9._-] collapse to _ to keep filenames safe.

#Fields

path is resolved relative to the process working directory. Use an absolute path in production (e.g. /var/log/snapr).

#Example

logs:
  path: /var/log/snapr
  system: true
  perJob: true
  maxSizeMB: 100
  maxBackups: 7
  maxAgeDays: 30
  compress: true

#Disabling files

Set system: false and/or perJob: false. With both off, no file sink is created and the API endpoints return:

503 Service Unavailable
{"error":"System log is disabled (logs.system=false)"}

The UI hides the log toolbar and shows a placeholder explaining which flag turned the panel off.

stdout is unaffected — journalctl, docker logs, or your container runtime still see everything.

#Edge cases

• Lumberjack rotation is in-process. A second snapr process pointed at the same path will fight for rotation; don't run two instances against one log directory.
• No external rotation needed. Don't combine snapr's rotation with logrotate(8)/copytruncate — set snapr's maxSizeMB: 0 if you'd rather an external tool own rotation.
• Disk full. Lumberjack swallows write errors silently to keep the process alive; the same event still hits stdout, so external collectors keep working.
• Removed jobs. When a job is deleted from the config, its jobs/<name>.log is left in place. Clean up manually if needed.
• Renamed jobs. A rename starts a new file under the new name; the old file is not migrated.
• server.enabled: false (scheduler-only mode). File logs still work the same way; only the API endpoints that read them are gone.
• Permissions. snapr creates path with 0755 and files with lumberjack's default (0600). Make sure the process user can write there — failure to create the directory aborts startup.