What is the difference between the various log messaging severity levels?

Question

There are several types of log messaging severity levels - emergency, critical, error, alert, info, notice, warning, debug. What does each level mean? Example, emergency and critical sound similar, what's the difference? Also, info and notice sound similar.

Answer 1

The code is the best manual :)

/**
 * Log message severity -- Emergency: system is unusable.
 */
define('WATCHDOG_EMERGENCY', 0);
/**
 * Log message severity -- Alert: action must be taken immediately.
 */
define('WATCHDOG_ALERT', 1);
/**
 * Log message severity -- Critical conditions.
 */
define('WATCHDOG_CRITICAL', 2);
/**
 * Log message severity -- Error conditions.
 */
define('WATCHDOG_ERROR', 3);
/**
 * Log message severity -- Warning conditions.
 */
define('WATCHDOG_WARNING', 4);
/**
 * Log message severity -- Normal but significant conditions.
 */
define('WATCHDOG_NOTICE', 5);
/**
 * Log message severity -- Informational messages.
 */
define('WATCHDOG_INFO', 6);
/**
 * Log message severity -- Debug-level messages.
 */
define('WATCHDOG_DEBUG', 7);

Answer 2

The levels are mostly arbitrary below alert, and just represent a general priority of "badness".

In theory, a minimum log level could be set up in core, so that you could enable logging on a live site, but only have certain levels logged (say, emergency, critical, error, and alert) and the rest ignored. To my knowledge, this hasn't been implemented (I would love to be shown wrong about this).

The current usage has more to do with filtering on the recent log messages page (admin/reports/dblog).