pcp
[Top] [All Lists]

Re: [pcp] Possible Bug in Linux PMDA

To: pcp@xxxxxxxxxxx
Subject: Re: [pcp] Possible Bug in Linux PMDA
From: Ken McDonell <kenj@xxxxxxxxxxxxxxxx>
Date: Mon, 13 Oct 2014 13:49:47 +1100
Delivered-to: pcp@xxxxxxxxxxx
In-reply-to: <1413164506.6110.15.camel@tardis>
References: <1413164506.6110.15.camel@tardis>
User-agent: Mozilla/5.0 (X11; Linux x86_64; rv:31.0) Gecko/20100101 Thunderbird/31.1.2
On 13/10/14 12:41, Chandana De Silva wrote:
$ pminfo -d -t disk.all.avactive

disk.all.avactive [total count of active time, summed for all disks]
     Data Type: 64-bit unsigned int  InDom: PM_INDOM_NULL 0xffffffff
     Semantics: counter  Units: millisec


Should the semantics be "instantaneous" ?

G'day Chandana.

Nope, this one is a counter.

If you sample it with pmval or pmchart or pmie then the "value" will be turned into a time utilization in the range 0 (no activity) to N.0 (for N (== hinv.ndisk) disks in the system, all devices are busy all of the time during the sample).

For practical purposes, one would probably choose disk.dev.avactive for the per disk utilization (in the range 0.0 to 1.0), else combine in pmie and divide by hinv.ndisk (to get the average disk utilization in the range 0.0 to 1.0).

<Prev in Thread] Current Thread [Next in Thread>