[mrtg] Logfile entries reset to zeroes after a few hours

[Louis van Dyk]

Hi Steve

Thank you for your reply.  

When I looked at the subjects of the e-mails coming back to root when
with all the MRTG errors, I noticed that there were two distinct subject
lines.

You were correct in the last statement in (b) ... there were TWO MRTG
processes being triggered.  I had not realised that Fedora 12 installs a
cron file as /etc/cron.d/mrtg.  I had ALSO created one in root's
crontab.  Since removing the one (root's crontab) I no longer am getting
errors in my mail, and so expect that the data will no longer be lost.

Oh, and I was running SELinix, and re-applied permissions, which
naturally made no difference.

Thanks for the assistance.  Perhaps you can add your answer to the FAQ
somewhere?  I saw a number of people asking the same question, but no
real answer being given.

Regards
Louis


On Sat, 2010-04-24 at 09:51 +1200, Steve Shipway wrote:
> All the values being reset to zero indicates that your .log file is
> being somehow deleted and re-created with no data.  My guess is one of
> two things.  Either :
>  
> a) You have some periodic job (eg, a 'tidy up' script) which is
> periodically removing the .log files for some reason.  Maybe a coding
> error or somesuch is causing the files to be incorrectly deleted.  To
> test for this, stop MRTG for a while, and see if the .log file
> disappear.
> b) The permissions or ACLs on the directory holding the .log files is
> such that the MRTG process has problems during the update process.
> Normally (I think!)  MRTG makes a copy of the .log file, updates this,
> then moves it back to replace the old one.  Possibly (from the errors
> you report) a problem occurs during this move/rename process.  Try
> doing a temporary chmod 777 on the directory and files to see if this
> fixes it (though this is not a good long-term solution of course).  Do
> you have SELINUX installed -- this could be affecting rights to remove
> files, though create and update is OK?  Do you accidentally have
> multiple MRTG processes running, such as one owned by root and one by
> a MRTG user?  And so on.
>  
> The erros indicate problems in renaming the files.  Check that there
> are not a load of *.tmp or *.old files hanging about in the directory
> (if so, delete them).  Also run an fsck on the filesystem, in case
> there is filesystem corruption affecting the directory.  
>  
> You might also like to consider moving to using MRTG with RRDTool
> instead of in Native mode (with .log files) as this is more powerful,
> albeit a bit more complex.
>  
> Steve
>  
> 
> ______________________________________________________________________
> From: mrtg-bounces+s.shipway=auckland.ac.nz at lists.oetiker.ch
> [mrtg-bounces+s.shipway=auckland.ac.nz at lists.oetiker.ch] On Behalf Of
> Louis van Dyk [louis at qdcec.co.za]
> Sent: Saturday, 24 April 2010 6:46 a.m.
> To: mrtg at lists.oetiker.ch
> Subject: [mrtg] Logfile entries reset to zeroes after a few hours
> 
> 
> 
> Hi
> 
> I have been running mrtg for many years on many different Linux boxes
> and have always had no problem with it.  But now I am in desperate
> need of assistance. 
> 
> I am running Fedora 12 and installed mrtg from the Fedora repository:
> mrtg-2.16.2-4.fc12.x86_64
> 
> I am only monitoring ONE device and have set the following entry in
> root's crontab file:
> */5 * * * *  env LANG=C /usr/bin/mrtg /etc/mrtg/mrtg.cfg
> 
> There are a number of interfaces being monitored on the device.  The
> CFG file was configured with cfgmaker.
> 
> THE PROBLEM:  Every few hours the logfile entries are reset to ZEROS
> through out the file, a few hours into the log file.  It is apparently
> random, in that if I look at the graphs they all cut off at different
> times over the last 24 hours.  So the graphs have only a few hours on
> each of them before flattening to zero.
> 

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.oetiker.ch/pipermail/mrtg/attachments/20100428/1a5a4ee5/attachment-0001.htm