Log Levels

Last modified 22 Apr 2021 17:31 +02:00

Log Levels

Level Circumstances Purpose Audience Example

FATAL

Critical errors. The system cannot continue operation, will crash or stop working

Detect critical system malfunction

System administrator, system management (e.g. Network Operation Center)

System out of memory

ERROR

Error that seriously affects the system, but the system as a whole can recover.

Detect system misconfiguration of major component failures

System administrator, Deployment engineer

Definition of resource c0c010c0-123456-…​ not found in the repository

WARNING

Suspicious situation. System may be able to operate normally, but there may be a hidden or temporary problem or an indication of future error.

Pro-active detection of problems

System administrator, Deployment engineer

Connection to host foobar timed-out, retrying in 10 seconds

INFO

Important changes in system state, start/stop of important system tasks, etc.

Rough system diagnostics (by non-developer)

Deployment engineer, Advanced system administrator

# MODEL # Enter addObject(ObjectContainer(AccountShadowType(null,Localhost OpenDJ-jack)))
Synchronization thread starting

DEBUG

Execution messages, state changes, expression evaluation messages and similar messages for system administrator.

Fine-grained system diagnostics (by non-developer)

Deployment Engineer

Executing outbound expression …​ for attribute "cn", resource "ds1", result: cpt. Jack Sparrow

TRACE

Fine-grained messages about execution details.

System diagnostics by the developer

Developer or advanced engineer

Moooo! Where the hell came the value of "foo" from?

Recommendations For Developers

Log Messages

Generaly speaking, avoid messages like:

  • Something bad happened somewhere, HA HA HA

  • FooTLA threw InvalidContextException while processing line 3324 of STDIN [ …​ 200 lines of stack strace …​ ] Caused by NullPointerException at MyThousandLinesLongClass:853.

Provide Meaningful Error Message

Provide all the information needed to diagnose the problem. Try to include root cause in the message.

Bad
Error resolving variable "foo"
Good
Error resolving variable "foo": Object not found (OID:123456-123456-...)

Do Not Use Developer’s Language

Try to think like a system administrator, not like a developer. It will not help administrator that an InvalidArgumentExeption was throws somewhere at the line 42 of MyDeeplyBurriedUtilClass. What administrator needs to know is that he has made a typo in the configuration file. The only log level that can use any kind of developer’s terminology is TRACE. Level DEBUG may use some geekish terminology, but should be generally understood by deployment engineers with a very limited knowledge of Java. The INFO level and above must not use any of the Java jargon.

Bad
IDMException was thrown at ProvisioningService:42, root cause: IllegalArgumentException at SchemaParser:32.
Good
Unexpected annotation "<foobar/>" while processing resource schema, attribute "cn", resource "ldap1"

Provide Context

The messages on WARNING, ERROR and FATAL levels should not depend on the context, they should stand alone and provide all the information needed. In other words, the administrator should not need to analyze 100 lines of log file to see where the error happened. The reason is that the log level of production systems is usually set to WARNING, therefore the message with lower log levels will not even appear in the log.

Bad
Error resolving a reference: Object not found
Good
Error resolving variable "foo" in outbound expression processing for variable "cn", resource "ldap42": Object not found (OID:123456-123456-...)
Was this page helpful?
YES NO
Thanks for your feedback