Permalink
Switch branches/tags
Find file Copy path
Fetching contributors…
Cannot retrieve contributors at this time
104 lines (60 sloc) 3.75 KB

How to remove spikes

.. index::
   pair: RRD; spikes;

If a plugins field is initially created with datatype COUNTER and no declaration of a .max value, there may occur spikes in the back-end RRD database when the counter goes back to zero. Here we show a simple way to fix this.

Set a MAX value

First, the plugins config must be enhanced to present a .max value.

Once this has been accomplished and the Munin master has polled the plugin at least once, the masters datafile will have the .max value stored for the plugin.

Then the following perl 1-liner can be used to update all matching RRDs. Replace DOMAIN with the name of the Munin domain being polled, and replace PLUGIN with the name of the plugins RRD database files that need updating.

cd /var/lib/munin/DOMAIN
perl -ne 'next unless /:PLUGIN/; if (/.*;(\S+):(\S+)\.(\S+)\.max\s+(\d+)/) {foreach (glob "$1-$2-$3-?.rrd") {print qq{File: $_\tMax: $4\n};qx{rrdtool tune $_ -a 42:$4};qx{rrdtool dump $_ > /tmp/rrdtool-xml};qx{mv $_ $_.bak};qx{rrdtool restore -r /tmp/rrdtool-xml $_};qx{chown munin:munin $_}}}' ../datafile

An example, updating the MAX value in all the xvm_ plugins that have been updated to report a .max value:

cd /var/lib/munin/example.com
perl -ne 'next unless /:xvm_/; if (/.*;(\S+):(\S+)\.(\S+)\.max\s+(\d+)/) {foreach (glob "$1-$2-$3-?.rrd") {print qq{File: $_\tMax: $4\n};qx{rrdtool tune $_ -a 42:$4};qx{rrdtool dump $_ > /tmp/rrdtool-xml};qx{mv $_ $_.bak};qx{rrdtool restore -r /tmp/rrdtool-xml $_};qx{chown munin:munin $_}}}' ../datafile
File: ic5.example.com-xvm_arch4-xvm_arch4_read_bytes-c.rrd    Max: 34359738352
File: ic6.example.com-xvm_arch3-xvm_arch3_read_bytes-c.rrd    Max: 34359738352
File: ic3.example.com-xvm_dwork-xvm_dwork_write_bytes-c.rrd   Max: 25769803764

Modify the RRD files concerned

Note

The following is a distillation of the process, outlined [1] by Greg Connor in 2004-05-06.

Here is a quick recipe for getting rid of spikes.

In this example dnscache is the name of the plugin and dns1.example.com is one of the affected hosts.

The desired MAX value is 1000.

Verify that the plugin now sets the right .max parameter. Fix the plugin if needed.

# cd /var/lib/munin
# cat datafile | grep dnscache | grep max

View the rrd file with "rrdtool info"

# rrdtool info DNS/dns1.example.com-dnscache-cache-c.rrd

to look for

ds[42].max = NaN

This indicates that the plugin didn't have a .max defined at the time the rrd file was started. If you don't mind losing the data, you can delete the RRD files at this point and the new ones will be created with the right max.

If you want to save the data, first modify each .rrd file so that it has a max for datasource 42 (not sure why it is always 42, probably a tribute to d.adams) (Best to do this right after an update like at :06 :11 or so)

# bash
# for j in `find /var/lib/munin -name "*dnscache*-c.rrd"`; do \
rrdtool tune $j -a 42:1000;
done

Note

Replace 1000 with the desired MAX value.

Finally, dump each rrd file to xml and restore with -r flag. There will be some output to let you know which data points were dropped and replaced with NaN.

# bash
# for j in `find /var/lib/munin -name "*dnscache*-c.rrd"`; do
rrdtool dump $j > /tmp/xml ;
mv $j $j~ ;
rrdtool restore -r /tmp/xml $j;
chown munin:munin $j ;
done

If you are impatient, rebuild one host's graphs and look at it, or just wait 5 min and check.

# su munin -c "/usr/share/munin/munin-graph --nolazy --host DNSOverview "

[1]See the post to munin-user mailing list by Greg Connor.