Skip to content

Logging

Selva uses Structlog for logging and provides some facilities on top of it to make its usage a bit closer to other frameworks like Spring Boot.

It is integrated with the standard library logging, so libraries that use it are logged through Structlog. It also enables filtering by logger name using the standard library.

Why?

Nowadays, it is very likely that your application is deployed to a cloud and its logs are sent to an aggregator like Graylog, so a structured logging format seems to be the logical choice.

For more information on why use structured logging, refer to the Structlog documentation.

Configure logging

Logging is configured in the Selva configuration:

logging:
  root: WARNING # (1)
  level: # (2)
    application: INFO
    selva: DEBUG
  format: json # (3) 
  setup: selva.logging.setup # (4)
  1. Log level of the root logger.
  2. Mapping of logger names to log level.
  3. Log format. Possible values are "json", "logfmt", "keyvalue"" and "console".
  4. Setup function to configure logging.

The format config defines which renderer will be used. The possible values map to:

value renderer
json structlog.processors.JSONRenderer()
logfmt structlog.processors.LogfmtRenderer(bool_as_flag=True)
keyvalue structlog.processors.KeyValueRenderer()
console structlog.dev.ConsoleRenderer()

If not defined, format defaults to "json" if sys.stderr.isatty() == False, or "console" otherwise. This is done to use the ConsoleRenderer during development and the JSONRenderer when deploying to production.

Manual logger setup

If you need full control of how Structlog is configured, you can provide a logger setup function. You just need to reference it in the configuration file:

logging:
  setup: application.logging.setup

import structlog
from selva.configuration import Settings


def setup(settings: Settings):
    structlog.configure(...)

The setup function receives a parameter of type selva.configuration.Settings, so you can have access to the whole settings.