Server Logging

The server logger can be configured to adjust the amount of detail that is included in the server logs. This can be helpful for debugging or troubleshooting server issues. You can change the verbosity level of the logger by setting the SERVER_VERBOSITY environment variable.

The SERVER_VERBOSITY variable accepts the following values: DEBUG, INFO, WARNING, ERROR and CRITICAL.

Server logging format

You can customize the format by setting the SERVER_LOGGING_FORMAT environment variable. By default it is set as SERVER_LOGGING_FORMAT=plaintext. If it is set to unknown format it will use default value instead.

Plaintext format

The plaintext logging format will be used by default. Or it can be set explicitly by setting the SERVER_LOGGING_FORMAT=plaintext environment variable.

By default, the server will log messages in the following format:


It can be customized by setting the SERVER_LOGGING_TEMPLATE environment variable. Example:

SERVER_LOGGING_TEMPLATE=$'%(asctime)s\t[%(name)25.25s]\t%(levelname)5s: %(message)s'


Format string with special characters like \t, \n, etc. can’t be passed through .env file as they are getting escaped.

To pass such string you can

  1. Use export SERVER_LOGGING_TEMPLATE=$'%(asctime)s\t[%(name)25.25s]\t%(levelname)5s: %(message)s' syntax for local python deployment or docker compose with exported env variables
  2. Use combination of .env file and -e flag for plain docker deployment docker run -d --env-file .env -e SERVER_LOGGING_TEMPLATE=$'%(asctime)s\t\t[%(name)25.25s]\t%(levelname)5s: %(message)s' mageai/mageai:latest

JSON format

To use JSON logging format, set SERVER_LOGGING_FORMAT=json.

JSON formatted logs will have the following fields:

  "timestamp": "...",
  "level": "...",
  "message": "...",
  "logger": "...",
  "module": "...",
  "function": "...",
  "line_number": "..."

More logging formats/customization to come…

Pipelines logging

By default, logs for each pipeline run will be stored in the <path_to_project>/pipelines/<pipeline_name>/.logs folder.

Learn more about logs here.

Set logging level

The default logging level is INFO. To customize the logging level for logs of block runs and pipeline runs, you can set the level of logging_config in your project’s metadata.yaml file.

Example logging config to only log errors:

  level: ERROR

Set log retention period

To delete old logs in your peristent volume, you can specify the retention_period in the logging_config. Valid period units are ‘h’, ‘d’, and ‘w’.

Example logging config:

  retention_period: '15d'

Then you can run command mage clean-old-logs [project_uuid] to clean up old logs

Block logging

A logger is available in the block’s context. You can use it to log messages to the block’s log file. The logger is passed into each block as a keyword argument. You can retrieve the logger by calling kwargs.get('logger') within the block.


def load_data(*args, **kwargs):
    kwarg_logger = kwargs.get('logger')'Test logger info')
    kwarg_logger.warning('Test logger warning')
    kwarg_logger.error('Test logger error')


Logging to external destination


To store logs in S3, you need to set the logging config in your project’s metadata.yaml file.

Example S3 logging config:

  type: s3
  level: INFO
    bucket: <bucket name>
    prefix: <prefix path>
    aws_access_key_id: <(optional) AWS access key ID>
    aws_secret_access_key: <(optional) AWS secret access key>
    endpoint_url: <(optional) custom endpoint url>

To authenticate with S3, the credentials need to be configured in one of the following ways:

  • Configure credentials in the logging_config.
  • Configure credentials in environment variables: AWS_ACCESS_KEY_ID and AWS_SECRET_ACCESS_KEY.
  • Authenticate with IAM role. Grant S3 access to the IAM role of the cloud instance.

The endpoint_url can be configured to connect to S3 compatible services (e.g. MinIO).

Google Cloud Storage

To store logs in GCS, you need to set the logging config in your project’s metadata.yaml file.

Example GCS logging config:

  type: gcs
  level: INFO
    path_to_credentials: <(optional) path to gcp credentials json file>
    bucket: <bucket name>
    prefix: <prefix path>

More destinations coming…

Edit pipeline logging

When you’re editing a pipeline (e.g. /pipelines/[uuid]/edit), you can execute the code for an individual block and see the output. Any print statements in the block of code is displayed in the block’s output.

However, you can redirect those print statements to output to logs.

To toggle this feature, go to the pipeline settings page (e.g. /pipelines/[uuid]/settings) and check the box labeled When running a block while editing a pipeline, output the block messages to the logs.