System

Tip

If you're running Plombery locally, in most cases you don't need to change these settings

allowed_origins

Change it if running in production.

It allows to configure the CORS header Access-Control-Allow-Origin, by default it's value is * so it allows all origins.

data_path

The absolute path to the data directory where logs and output data is stored.

By default is set to the current working directory.

database_url

The DB URI, by default sqlite:///./plombery.db

database_auth_token

The auth token for libsql databases hosted on Turso cloud

blocked_loop_threshold

Plombery runs a task declared with async def on the event loop, and a task declared with def in a thread. A coroutine that calls blocking code — a plain time.sleep, a synchronous HTTP client, a long CPU loop — holds the loop for its whole duration, and while it does nothing else in the process runs: the API stops answering, the live logs stop streaming and the scheduler stops firing. The symptom is a frozen UI with nothing in the logs to explain it.

Plombery watches for this and logs a warning naming the tasks that were running:

The event loop was blocked for 12.4s while these tasks were running:
fetch_report. An async task must not call blocking code: use await, or declare
the task with `def` instead of `async def` so that Plombery runs it in a thread.

The threshold is 2 seconds by default. Set it to 0 to turn the check off.

plombery.config.yaml

blocked_loop_threshold: 2

Tip

The fix is almost always one of two things: await the blocking call if an async version exists, or drop the async from the task definition, so that Plombery runs it in a thread where blocking is harmless.

retention

By default Plombery keeps every run forever. Set one or both thresholds to reclaim space automatically; the policy is applied at startup and then once a day.

plombery.config.yaml

retention:
  # delete the log files of runs that finished more than 30 days ago,
  # keeping the runs themselves so the history and the charts are intact
  files_days: 30
  # delete the runs that finished more than a year ago, with their task
  # runs, their stored outputs and their log files
  runs_days: 365

Both are optional and independent: setting only files_days keeps the whole history while dropping the logs, setting only runs_days deletes old runs entirely.

Note

Task outputs are stored in the database, not on disk, so runs_days is the threshold that reclaims most of the space. files_days only affects log files.

Runs that haven't finished are never deleted, however old they are, and data directories left behind by runs that are no longer in the database are cleaned up as well.

frontend_url

The URL of the frontend, by default is the same as the backend, change it if the frontend is served at a different URL, for example during the frontend development.

pipeline_version

The version recorded on every run, for the pipelines that don't declare a version of their own.

Set it to the revision the deployment was built from — as an environment variable, PIPELINE_VERSION=$(git describe --tags --always) — since an image ships without the git history Plombery would otherwise read the version from.

When it's unset and there is no repository either, runs record no version.