[Bug 919429] Re: Munin "IO Service Time" graph gives completely implausible numbers

Tel lists at lnx-bsp.net
Fri Jan 20 23:32:46 UTC 2012


I can see one problem in the file /usr/share/munin/plugins/iostat_ios as
follows:


LINE 202:             print("${dev}_rtime.value ", ($rtime_diff != 0) ? ($rio_diff / $rtime_diff) : 0, "\n",
LINE 203:                   "${dev}_wtime.value ", ($wtime_diff != 0) ? ($wio_diff / $wtime_diff) : 0, "\n",
LINE 204:                   );


Here we see it divides number of IO operations by time in milliseconds. However the graph has the vertical axis in seconds. Dividing by time gives Hz never back to seconds again. So the graph results are really in Hz (not a useful unit given the context we are working with here). I suggest that the calculation should be:


LINE 202:             print("${dev}_rtime.value ", ($rio_diff != 0) ? ($rtime_diff / $rio_diff) : 0, "\n",
LINE 203:                   "${dev}_wtime.value ", ($wio_diff != 0) ? ($wtime_diff / $wio_diff) : 0, "\n",
LINE 204:                   );


There's another (minor) problem which is to say that returning 0 in a situation where no IO has occurred is a lie, should return NaN or NA but if you read the Munin protocol specification (see link below) it claims "Output must be integer or decimal number," so the have no provision for a plugin saying "hey, this value does not exist right now" and that's strange because the RRD system does support NaN for missing values -- but that's a bigger problem for another day.

http://munin-monitoring.org/wiki/protocol-config

-- 
You received this bug notification because you are a member of Ubuntu
Server Team, which is subscribed to munin in Ubuntu.
https://bugs.launchpad.net/bugs/919429

Title:
  Munin "IO Service Time" graph gives completely implausible numbers

To manage notifications about this bug go to:
https://bugs.launchpad.net/ubuntu/+source/munin/+bug/919429/+subscriptions



More information about the Ubuntu-server-bugs mailing list