Hi,
Thanks for the explanation that makes it perfectly clear. It is actually flapping green/white/green/white/green....
I'll have to have a look at the script to figure out why. That should take away a lot histlogs.
- Roland
From: Japheth Cleaver [cleaver at terabithia.org] Sent: Friday, 20 December 2013 12:46 AM To: Roland Soderstrom Cc: xymon at xymon.com Subject: Re: [Xymon] histlogs cleanup
On Wed, December 18, 2013 3:00 pm, Roland Soderstrom wrote:
Hi,
I got this xymon ext script checking a Sun StorageTek 6180 array. It works quite well, each run generates a lot of status and data. The problem is it generates about 35000 files in Xymonserver: xymon/data/histlogs/a6180/<test> a month.
I'm very interested in keeping history as in graphs and alerts. What are these histlogs files used for by the xymonserver? If they are used just to generate things as they come in then it's no problem deleting them every 2 weeks or so. But if they are used to generate graphs or see alerts history etc then you really don't want to delete them
The "histlogs" directory is used to store snapshots of status messages whenever there's a threshold or "color" change. If you drill down on the history page for a given host+test combo, you can click the color dot at each timestamp to see what the output or test result was at that moment (well, when that color/status was first sent for that time period).
35000 files in a host+test directory in a month indicates that, on average, your test is changing colors 1100/day, which is suspiciously close to 1440/day, or once a minute.
If you look at the http://your.server.here/xymon-cgi/history.sh?HISTFILE=a6180.<test> page, is it flapping green->yellow/red->green (etc)? If so, you might want to look at the thresholds and logic calculation in the originating script and verify that it's doing what you want it to do.
If it's going green->purple->green (etc), it's possible you're manually setting the TTL on the status message too low for the frequency at which the script is running.
But no, if your script is feeding in to xymond_rrd and the graph display is actually being handled by the RRD subsystem, then deleting status history logs won't affect graphing. It *will* mean, that you won't be able to view a "snapshot" of what the alert was at that moment in time; you'll just have a record that the color, in fact, went red/yellow, at that point. (Those records are kept as single lines in per-test files in the /hist/ directory.
HTH,
-jc