clearing crc error counter on switch leads to huge numbers in crc error graph

mark

Hi, we just cleared the crc error counter on a switch, and it seems that stor2rrd misinterprets the negative number (calculated "current errors=0" - "number of last errors before clearing>0") as integer (and not as unsigned integer or something like that). The graph showed us a huge number for the crc errors (millions). I think if the result is a negative number it should be set to 0 by stor2rrd.

Pavel

Hi,

this might happen if you clear the counter, there is no way which would avoid it completly if this happens.

use this method to clean the peak out.
get script, gunzip and copy to /home/stor2rrd/stor2rrd, http://www.stor2rrd.com/download/remove_spikes_from_rrdtool_single.sh.gz


Usage example is for SAN switch ASAN11 and port 14 (note file has index 15 if you use stor2rrd 1.30, for lower stor2rrd use 14) : data/ASAN11/port15.rrd
Column name is: swFCPortRxCrcs
data limit 1000 (peaks higher than 1000 errors per seconds will be removed)

# su - stor2rrd
$ cd /home/stor2rrd/stor2rrd
$ ksh ./remove_spikes_from_rrdtool_single.sh
Provide full path to RRDTool file
data/ASAN11/port15.rrd

Provide column name
swFCPortRxCrcs

Provide data limit
1000

Working for: data/ASAN11/port15.rrd:swFCPortRxCrcs and limit 1000
Actual MAX limit of that item in RRDTool file is:
Load has been done with MAX limit: 1.0000000000e+03


I hope it is clear and will work for you

Pavel

mark

Hi Pavel,

thanks for the quick reply - yes, this works for me, thanks!

Just thought there might be a way to avoid those peaks all together...

Pavel

It is hard to avoid such peaks when the source is type counter.