Clarification on log rotation

Clarification on log rotation

[Andreas Hasenack]

Hi,

I'm checking auditd's native logrotation mechanism.

The auditd.conf manpage states this for num_logs:

"The excess log check  is  only  done  on startup and when a
reconfigure results in a space check."

I kept generating events, and truth be told, no rotation happened once
the logfile size was above max_log_file. At least not after a few
minutes.

When does a space check happens, besides on a restart? Just external
events likg SIGUSR1 and perhaps SIGHUP?

Since these are external events, how do sysadmins deal with log
rotation: completely ignore auditd's native mechanism and setup
logrotate as usual?

--
Linux-audit mailing list
Linux-audit@redhat.com
https://www.redhat.com/mailman/listinfo/linux-audit

Re: Clarification on log rotation

[Steve Grubb]

On Monday, November 23, 2020 9:21:56 AM EST Andreas Hasenack wrote:
> I'm checking auditd's native logrotation mechanism.
> 
> The auditd.conf manpage states this for num_logs:
> 
> "The excess log check  is  only  done  on startup and when a
> reconfigure results in a space check."
> 
> I kept generating events, and truth be told, no rotation happened once
> the logfile size was above max_log_file. At least not after a few
> minutes.

Rotation is different than excess log checks. Log size checking is done every 
write. But this is only done when the daemon is not in debug mode and  
write_logs is not 0 and max_log_size_action is rotate and num_logs > 1.

> When does a space check happens, besides on a restart? Just external
> events likg SIGUSR1 and perhaps SIGHUP?

Every 3 writes.

> Since these are external events, how do sysadmins deal with log
> rotation: completely ignore auditd's native mechanism and setup
> logrotate as usual?

Generally people fall into 3 camps. The first camp is they correctly configure 
the native implementation and just use it. The second camp need something 
special. They either set max_log_size_action to keeplogs and then handle it 
on a cron job where that may use checkpointing. And yet another group just 
sends events to syslog and handle it via splunk or elastic search.

-Steve


--
Linux-audit mailing list
Linux-audit@redhat.com
https://www.redhat.com/mailman/listinfo/linux-audit