T.J. Yang wrote:
This is a write up of my recent effort to debug and understand error message in rrd-[status|data].log files.
What is the problem ?
xymon(hobbit) client can send two pairs of data at the same time stamp.
"/usr/bin/kstat -p -s '[or]bytes64'| sort" will give out wrsmd and mac info on a solaris 10 box.
and causing bge:1:mac:obytes64 try to overwrite bge:0:mac:obytes64, thus the error of "minimum one second step"
1. with wrsmd filtered out only
hobbit at test$ /usr/bin/kstat -p -s '[or]bytes64'|egrep -v wrsmd | sort
bge:0:bge0:obytes64 50828816355
bge:0:bge0:rbytes64 81490514060
bge:0:mac:obytes64 50828816355
bge:0:mac:rbytes64 81490514060
bge:1:mac:obytes64 0
bge:1:mac:rbytes64 0
bge:2:mac:obytes64 0
bge:2:mac:rbytes64 0
bge:3:mac:obytes64 0
bge:3:mac:rbytes64 0
2. My Solution, filtering out both wrsmd and mac. let me know if you have better one.
hobbit at test$ /usr/bin/kstat -p -s '[or]bytes64'|egrep -v 'wrsmd|mac' | sort
bge:0:bge0:obytes64 50829477390
bge:0:bge0:rbytes64 81501640315
hobbits at test$
We have been applying the same kind of fix for Xymon clients running on recent Sun hardware like T2000, X4450 or M4000. Beside the "minimum one second step" issue, useless graphs for "wrsmd" and "mac" will be created.
Henrik, can you include this changes in the source tree ?
Dominique