Linode CPU stats on cacti

My usual first guess is always scale mismatch: the Linode CPU graph ranges from 0 to 400% while some tools always scale it to 0-100%, but the graphs don't differ by a factor of four so that might not be it.

I might try running a single CPU-bound thread for awhile (~30 minutes or so), spinning one CPU at ~100%… throwing a big, constant load at both graphs simultaneously should establish what sort of error it is.

Just out of curiosity, which kernel/version are you running? Ultimately, the number cacti is reporting comes from the kernel, so perhaps there's a known (or at least reproducible) difference between that kernel version and whatever the hypervisor-dowsing expedition feeds Linode's graphs.

Thanks, running one core at 100% showed me which cpu graphs were incorrect. I used a different 4 core template (Host MIB) and now it matches linodes.

@linodeTail:

I used a different 4 core template (Host MIB) and now it matches linodes.
Care to share your results so if someone else has the same problem in the future?

–

Travis

I used verbatim the 4 cpu template described here:

http://www.goldfisch.at/knowwiki/howtos … u_graphing">http://www.goldfisch.at/knowwiki/howtos/cacti#multi-cpu_graphing

pictutorial here:

http://www.powercram.com/2010/01/multi- … cacti.html">http://www.powercram.com/2010/01/multi-cpu-graphing-in-cacti.html