[asterisk-users] queue show <queue-name> vs queue log for calculating average hold time

[Paul Belanger]

On Wed, Jan 28, 2015 at 1:37 PM, Paul Belanger
<paul.belanger at polybeacon.com> wrote:
> On Wed, Jan 28, 2015 at 12:23 PM, Ishfaq Malik <ish at pack-net.co.uk> wrote:
>> Hi
>>
>> We're using 1.8.23.1 on CentOS 5 and are trying to get accurate stats for
>> queues.
>>
>> For a particular customer, when I run queue show <queue_name> I get the
>> following numbers:
>>
>> <queue_name> has 0 calls (max unlimited) in 'ringall' strategy (17s
>> holdtime, 94s talktime), W:0, C:175, A:44, SL:48.6% within 45s
>>
>> So from that data we look at
>> 17s holdtime
>> And assume that is the average hold time before calls get answered by a
>> queue members.
>>
>> However, if I calculate the average hold time from out queue log table using
>> the following SQL
>>
>> select sum(data1)/ count(*) as ave_hold_time from queue_log where time >
>> DATE(NOW()) and queuename='<queue_name>' and event='CONNECT';
>>
>> I get the vastly different figure of 92.4.
>>
>> So, is the queue show figure wrong due to a bug or am I making an incorrect
>> assumption as to what it means?
>>
>> Thanks in advance
>>
> Welcome to business logic embedded into app_queue.  The issue with the
> queue show command rendering stats, is what timeframe are the stats
> aggregated over?  IIRC, the calculations are using a moving
> average[1].
>
Opps, sent instead of pasting.

Either way, your likely better off rendering the data using the raw
sql info vs depending on CLI output.  That's what we've done.

[1] http://en.wikipedia.org/wiki/Moving_average
-- 
Paul Belanger | PolyBeacon, Inc.
Jabber: paul.belanger at polybeacon.com | IRC: pabelanger (Freenode)
Github: https://github.com/pabelanger | Twitter: https://twitter.com/pabelanger