---

From: Jay Leafey <jay.leafey@mindless.com>
To: CentOS mailing list <centos@centos.org>
Sent: Wednesday, February 11, 2009 9:40:30 PM
Subject: Re: [CentOS] logs such as messages, boot.log, and kernel contained 0 size

Frank Ling wrote:
> Hi,
>
> My both CentOS 5 servers have logging problems. Logs such as messages, boot.log, kernel, spooler, and tallylog in /var/log directory are all 0 size.
>
> The kernel is:  Linux 2.6.18-92.1.22.el5 #1 SMP.
>
> Since the /var/log/messages contained no information it would be impossible to troubleshoot the problem.
>
> I am very sure both systems have not been hacked by others.
>
> Sincerely,
>
> Frank Ling
> --------------------------------------------------------------------------
> -rw-------  1 root        root          0 Feb  8 04:02 messages
> -rw-------  1 root        root          0 Feb  3 11:04 messages.1
> -rw-------  1 root        root          0 Jan 25 04:02 messages.3
> -rw-------  1 root        root          0 Jan 11 04:03 messages.4
> -rw-------  1 root        root        10 Dec 27 13:00 messages.offset
>
> -rwx------  1 root        root          0 Feb 11 19:12 kernel
> -rwx------  1 root        root          0 Feb 11 16:53 kernel.1
> -rwx------  1 root        root          0 Jan 25 04:02 kernel.3
> -rwx------  1 root        root          0 Jan 11 04:03 kernel.4
>
> -rw-------  1 root        root          0 Feb  8 04:02 spooler
> -rw-------  1 root        root          0 Feb  3 07:51 spooler.1
> -rw-------  1 root        root          0 Jan 25 04:02 spooler.3
> -rw-------  1 root        root          0 Jan 11 04:03 spooler.4
>
> -rw-------  1 root        root          0 Jun 24  2008 tallylog
> --------------------------------------------------------------------------
>

I've had something similar happen a couple of times after an update.  In my case the /etc/services file got it's security context clobbered when some package tried to update it's contents.  When logrotate ran, the syslog daemon couldn't open /etc/services because of the error and I ended up with a bunch of empty log files.

The quickest way to check for this is the command:

    restorecon -v /etc/services

If nothing prints out in response, that's not the problem.  If it DOES, that might explain it.  I have been checking the contexts occasionally to try and trap exactly when it happens.  I use:

    restorecon -R -n -v /etc

which walks through the entire /etc tree looking for contexts to change but just reports any exceptions.

Just a thought!
-- Jay Leafey - Memphis, TN
jay.leafey@mindless.com