[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Another RRD question



All,

Somewhat similar question to one asked earlier about custom data/graphing

I have a script that gets SAR data. I am creating the rrd files manually up
front, as they have a different
step frequency that hobbit would create.

It outputs (for example)

cpu0pusr:number
cpu0pwio:number
cpu0pidl:number
cpu1pusr:number
cpu1pwio:number
cpu1pidl:number

etc...
This can vary based on the number of cpu's in the box. I am only doing this
on a limited number of hosts (3) so
hand creating the RRD's isn't too big of a deal. The problem is graphing.
Some of these have 16 CPU's, some only 4.
If I create a graph definition, it would have to be universal (i.e have up
to 16 definitions for everything) but that would
make the graphs for the boxes with only 4 CPU's not look right (I think,
havn't tested yet)

What I was wondering is, is there a way in the hobbitgraph.cfg to do
something like:
(NOTE: I know this syntax isn't correct, it's more for readability purposes)

[sar]
        TITLE sar data
        YAXIS percent
        DEF:cpu$NUMpusr=sar.rrd:cpu$NUMpusr:AVERAGE
        DEF:cpu$NUMpwio=sar.rrd:cpu$NUMpwio:AVERAGE
        DEF:cpu$NUMpidl=sar.rrd:cpu$NUMpidl:AVERAGE
        LINE1:cpu$NUMpusr#CC3333:cpu$NUM percent usr
        LINE1:cpu$NUMpwio#FF0000:cpu$NUM percent wio
        LINE1:cpu$NUMpidl#0000FF:cpu$NUM percent idle
        COMMENT:\n
        GPRINT:cpu$NUMpusr:LAST:CPU $NUM usr \: %5.1lf%s (cur)
        GPRINT:cpu$NUMpusr:MAX: \: %5.1lf%s (max)
        GPRINT:cpu$NUMpusr:MIN: \: %5.1lf%s (min)
        GPRINT:cpu$NUMpusr:AVERAGE: \: %5.1lf%s (avg)\n
        GPRINT:cpu$NUMpwio:LAST:CPU $NUM wio. \: %5.1lf%s (cur)
        GPRINT:cpu$NUMpwio:MAX: \: %5.1lf%s (max)
        GPRINT:cpu$NUMpwio:MIN: \: %5.1lf%s (min)
        GPRINT:cpu$NUMpwio:AVERAGE: \: %5.1lf%s (avg)\n
        GPRINT:cpu$NUMpidl:LAST:CPU $NUM idle. \: %5.1lf%s (cur)
        GPRINT:cpu$NUMpidl:MAX: \: %5.1lf%s (max)
        GPRINT:cpu$NUMpidl:MIN: \: %5.1lf%s (min)
        GPRINT:cpu$NUMpidl:AVERAGE: \: %5.1lf%s (avg)\n


I think the more correct way is to split the RRD files into sar0.rrd,
sar1.rrd, etc... and use the FPATTERN thing with the index.
I can split the output up and change the test name for each cpu to sar0,sar1
etc.. so that it updates the correct
RRD, but I just wanted to find out first if there is a way to do this in 1
RRD and do the work in the graph portion.

Thanks,
Jeff