Bug 1324248

Document URL: 

Administration guide
Command and configuration guide


Describe the issue: 

1.3.5 is now able to host multiple logging backends. This is controlled by the value in cn=config, nsslapd-logging-backend. Invalid values to this are ignored. Valid values are:

dirsrv-log
syslog
journald


Log events (regardless of activation) will be logged to subsystems in this order:

dirsrv-log
syslog
journald.

This is so that the highly reliable dirsrv-log backend is guaranteed to get the message first, followed by the remaining backends. 

There are some warnings about journald specifically that we MUST document. See this comment: https://fedorahosted.org/389/ticket/47968#comment:9

However the summary:

Use of journald will cause:


    Background consumes +15% cpu on an i7 system.
    Consumes 1200MB / hour for access / error logs.
    Consumes 1740MB / hour for audit logs. 

Additionally, Journald logs are hardcoded maximum to 4GB, which means at a busy site, ~3 hours or LESS of logs can be stored before the entire journal is rotated and logs lost! 

As well, journald rate limits busy inputs, and will drop ~85% of all slapd events without reconfiguration to prevent this.


https://fedorahosted.org/389/ticket/47968