[Dirvish] Too much disk space for images
jasonb at edseek.com
Fri Oct 2 18:55:18 UTC 2009
Paul Slootman wrote:
> On Fri 02 Oct 2009, Eric Smith wrote:
>> While I definitely agree that looking at the logfiles is helpful,
>> remember that an image (call it C) might have contained few delta files
>> when it was created, because the files were in a previous image B, but
>> not in an even earlier image A. If B is deleted before A, then C will
>> now hold the only reference to those files it had in common with B,
>> giving C a larger image size (as reported by du) than when it was
>> created (as reported in the logfile).
> So, be sure to check for unexpectedly large incrementals every day :-)
I've been meaning to write a munin plugin that graphs that every day for
about five years now. Obviously, I still have not. ;)
The challenge is munin runs every 5 minutes, so I'd need the plugin to
only actually run `du` on the latest snapshot per vault every 24 hours
and return the cached size the rest of the interval. Not particularly
difficult, but it would take longer than writing this email... or I'd
just do it.
Also been mulling a switch to Nagios for about 5 years. ;)
Maybe when I deploy chef in about 5 years...
Actually, I could probably just use the log.gz file. Clearly never
taken this beyond the "I probably should..." stage.
More information about the Dirvish