[rrd-users] Core dumped on graph generation/OpenBSD

[tichodroma at free.fr]

Hello.

First of all, kudos for the great soft, which I have been using for a  
while now.

Now for the bad news : I have rrdtool *sometimes* spitting cores while  
generating certain graphs (datas covering last year). The generation  
of other graphs from the same rrd works OK (datas representing last  
month, week, day and hour). I can reproduce the same problem on two  
platforms.

Setup :
- platform 1 : OpenBSD 4.6 GENERIC#5 i386, rrdtool-1.2.30
- platform 2 : OpenBSD 5.2 GENERIC.MP#368 amd64, rrdtool-1.2.30p3
- rrd generated by :
   rrdtool create /var/db/perfs_reseau-holding.rrd \
   --step 60 \
   DS:PktLossSrvGipssi:GAUGE:120:0:100 \
   DS:LatenceSrvGipssi:GAUGE:120:0:10000000 \
   DS:PktLossIPPublique:GAUGE:120:0:100 \
   DS:LatenceIPPublique:GAUGE:120:0:10000000 \
   RRA:AVERAGE:0.5:1:61 \
   RRA:AVERAGE:0.5:2:1442 \
   RRA:AVERAGE:0.5:14:10094 \
   RRA:AVERAGE:0.5:62:44702 \
   RRA:AVERAGE:0.5:730:526330
- rrd transfered from one platform to the other by export/import.
- rrd updated by :
   rrdtool update /var/db/perfs_reseau-holding.rrd --template  
PktLossSrvGipssi:LatenceSrvGipssi:PktLossIPPublique:LatenceIPPublique  
N:0:$LatenceSrvGips\
si:0:$LatenceIPPublique

Action leading to problem, on both platforms : generation of a graph  
covering a year of data, every minute
   rrdtool graph /var/www/stats_reseau/ping-gipssi-annee.png \
   -w 400 -h 200 -a PNG \
   --slope-mode \
   --start end-1y --end now \
   --title "Liaison avec holding - Latence" \
   --watermark "$(date +%a\ %d/%m/%Y\ %H\h%M\'%S\")" \
   --vertical-label "latence (ms)" \
   --units-exponent 0 \
   --lower-limit 0 \
   --color CANVAS#123 \
   --color BACK#658b98 \
   HRULE:70#00FF00:"Seuil OK" \
   HRULE:250#FF8800:"Seuil NON OK\n" \
    
DEF:LatenceSrvGipssi=/var/db/perfs_reseau-holding.rrd:LatenceSrvGipssi:AVERAGE  
\
    
DEF:LatenceIPPublique=/var/db/perfs_reseau-holding.rrd:LatenceIPPublique:AVERAGE  
\
   LINE1:LatenceSrvGipssi#FFFFFF:"Serveur GIPSSI (ms)" \
   LINE1:LatenceIPPublique#FFFF00:"Modem holding (ms)\n" \
   GPRINT:LatenceSrvGipssi:LAST:"En cours\t%5.0lf" \
   GPRINT:LatenceIPPublique:LAST:"\t%5.0lf\n" \
   GPRINT:LatenceSrvGipssi:AVERAGE:"Moy.\t\t%5.0lf" \
   GPRINT:LatenceIPPublique:AVERAGE:"\t%5.0lf\n" \
   GPRINT:LatenceSrvGipssi:MAX:"Max.\t\t%5.0lf" \
   GPRINT:LatenceIPPublique:MAX:"\t%5.0lf\n" \
   GPRINT:LatenceSrvGipssi:MIN:"Min.\t\t%5.0lf" \
   GPRINT:LatenceIPPublique:MIN:"\t%5.0lf"

Result : Segmentation fault      (core dumped)

I found a similar message in the list  
(http://www.mail-archive.com/rrd-users@lists.oetiker.ch/msg02760.html),  
without answer. Given that the problem occurs not allways, but from  
time to time, could it be related to some missing datas in the RRD ? I  
also found messages for similar problems (Core dump on large step  
sizes). I'm stucked, though.

Either, did I make a mistake in the RRD creation ?

What infos could be valuable in the core file to any of you willing to  
have a look at it ?

Thanks a lot.

Olivier Debre