[Sussex] Fedora Core 5 - lossing disk space

[Nico Kadel-Garcia]

On 1/31/07, Brendan Whelan <b_whelan at mistral.co.uk> wrote:
>
> Vic,
>
> Yes, I did delete the files as each was over 10GB and I needed to get the
> system back in action . However, I then trapped the first entries in the
> newly created log file and fixed the cause of the excess messaging.


There's a booby trap there: if you don't restart the process that is
generating files of doom, they won't necessarily get deleted until you
restart the process generating them (which has to do with the fopen and
unlink functions in C, look them up if you're curious, but it's not an
uncommon problem in dealing with mongo large log files). It's often good to
have a log rotation tool in place, and for server class machines a bit of
monitoring of over-filled file systems. (That sort of thing is what Nagios
and SNMP and MRTG were designed for!)

The daily cron jobs that are part of most Linux distributions are very good
for this, and can be easily modified to support new types of log files: I
consider log rotation to be a vital part of building the full software
package for any daemons I manage, and it can be helpful to get  a log
analysis utility like LogWatcher configured to look for whackiness like you
describe in the log files, so that you get a nightly report of "we saw
10,000,000 entries just like this".

What was it, anyway?
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mailman.lug.org.uk/pipermail/sussex/attachments/20070131/36eb469b/attachment.htm