Subject: Re: rrdtool dumps core
To: NetBSD/pkgsrc users <pkgsrc-users@NetBSD.org>
From: Martti Kuparinen <martti.kuparinen@iki.fi>
List: pkgsrc-users
Date: 10/31/2007 14:17:44
Martti Kuparinen wrote:
> I'm using rrdtool-1.2.23nb1 on several NetBSD servers (i386 3.1.1 and
> amd64 4.0_RCx) and on all servers I see weird crashes (=core dump) when
> generating the "last 12 months" image if the time is 03:00 - 04:55.
>
> During other times (05:00-24:00-02:55) I have no problems at all. Also I
> have no problems generating the "last 24 hours"/"last 7 days"/"last 4
> weeks" images even if the time is between 03:00 and 04:55.
After switching from summer time to normal time the core dumps started to appear
at 02:00 (AM). After some debugging and help from the rrdtool author, the cause
of this problem was identified :-)
Basically this was EDUMMYUSER (=dummy user error), the RRA statements for my
"year statistics" were saving history data for 288 years (!) instead of the
intended period of 1 year and according to the author this crash most likely has
something to do with time_t overflow. Why this only happens when I'm running
rrdtool at UTC midnight (or 02:00 in my local time) is still unclear to me...
Anyway, pay attention to your RRA lines and read the man page :-) This is what I
have now with our 48+2 port HP ProCurve switches (with 5 year of history):
MAXINT=4294967295
${RRDTOOL} create ${1} -s 300 \
DS:i1:COUNTER:600:0:${MAXINT} \
DS:o1:COUNTER:600:0:${MAXINT} \
...
DS:i50:COUNTER:600:0:${MAXINT} \
DS:o50:COUNTER:600:0:${MAXINT} \
\
RRA:AVERAGE:0.5:1:288 \
RRA:MIN:0.5:1:288 \
RRA:MAX:0.5:1:288 \
\
RRA:AVERAGE:0.5:6:336 \
RRA:MIN:0.5:6:336 \
RRA:MAX:0.5:6:336 \
\
RRA:AVERAGE:0.5:24:372 \
RRA:MIN:0.5:24:372 \
RRA:MAX:0.5:24:372 \
\
RRA:AVERAGE:0.5:288:1830 \
RRA:MIN:0.5:288:1830 \
RRA:MAX:0.5:288:1830
Martti