[collectd] question about new "disk" metrics

Marek Becka marek.becka at superhosting.cz
Fri Jul 3 16:52:47 CEST 2015

On 06/16/2015 11:12 AM, Markus "Shorty" Uckelmann wrote:
> Am 16.06.2015 um 06:32 schrieb Markus "Shorty" Uckelmann:
>> Am 16.06.2015 um 00:20 schrieb Gerardo Herzig:
>>> Hi all. Im trying to use the "disk" metrics of the new 5.5
>>> colletcd version, in order to avoid the "iostat" ruby wraper.
>>> What is cool about "iostat -x" is the "disk util%", to knowing
>>> how stressed the disk are.
>>> So, it is not clear to me what maths are involved in that
>>> calculations. Anyone knows if it is possible now with the new
>>> disk-plugin metrics?
>> Under Linux it should (sorry, haven't checked the code) derive from
>> the iostats utility. From the man page:
>> %util Percentage  of  elapsed time during which I/O requests were
>> issued to the device (bandwidth utilization for the device). Device
>> saturation occurs when this value is close to 100% for devices
>> serving requests serially.  But for devices serving requests in
>> parallel, such as RAID arrays and modern SSDs, this number does not
>> reflect their performance limits.
> Mea culpa! I just didn't read your post correctly. I can't answer your
> questions.
> So far I was able to find the "weighted_time" variable in the code of
> disk.c. This should be the last field from /proc/diskstats which
> contains the utilization as described above. But I'm not sure what
> happens to this metric (my C is bad). Because it doesn't show up in
> the output of the plugin.
> So someone with knowledge of that plugin might jump in. I would be
> interested in the disk utilisation too ;)
> BTW: Some interesting links I found:
> http://www.mjmwired.net/kernel/Documentation/iostats.txt
> http://www.xaprb.com/blog/2010/01/09/how-linux-iostat-computes-its-results/ 
> https://stackoverflow.com/questions/4458183/how-the-util-of-iostat-is-computed 

Hi, I'm using own patch for disk plugin that adds disk utilization and 
number of in-flight requests.  I didn't request pull to the upstream 
yet. The patch can be found at https://github.com/marekbecka/collectd/pull/1

More information about the collectd mailing list