Open Stack

No one needed to monitor Keystone throughput and response times ?

BUMP!

*
*
*
*
*Alejandro Comisario
#melicloud CloudBuilders*
Arias 3751, Piso 7 (C1430CRG)
Ciudad de Buenos Aires - Argentina
Cel: +549(11) 15-3770-1857
Tel : +54(11) 4640-8443



On Mon, Dec 3, 2012 at 5:01 PM, Alejandro Comisario <
alejandro.comisario at mercadolibre.com> wrote:

> Hi everyone.
>
> We have a production Keystone (Essex 2012.1.3) pool composed of 10 servers
> reading to the same database ( running MySQL Galera Cluster ).
> We have other Openstack part of the infrastructure like swift monitored
> over NewRelic ( python client )
>
> The thing is that we are trying to monitor keystone ( we want to know the
> overall RPM and ResponseTime of the entire pool ) and we are trying to do
> that with NewRelic.
> We are decorating the function "__call__" on the class "Application" on
> "keystone.common.wsgi" since every Keystone service inherits this class
> to link URLS, the servers appears in newRelic but we seen response times
> of 20.000 ms are a little bit higher ( when we know the response time are
> 200ms ) and few rpm per server (100 vs 400).
>
> Few questions :
>
> #1 Is ok to decorate this class ? is that class the main endpoint of each
> keystone request to capture all api calls ?
> #2 Dows anyone has any experience in monitoring kesytone with NewRelic ?
> #3 In case you are monitoring this metrics, what do you use to effectively
> measure the total pool RPM's and ResponseTimes ?
>
> Waiting for advices !
> Cheers !
> *
> *
> *---------------*
> *Alejandrito*
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstack.org/pipermail/openstack-operators/attachments/20121205/cc7bdb58/attachment.html>