RE: [tosca] monitoring

Hi Luca,

In our monitoring group, we discussed several use cases that we can address. One of them was the ability to define custom metrics, possibly using endpoints.

As a first stage, we decided to focus on a “minimal use case” with a few predefined metrics. We are now trying to finalize a design and yaml definition for this minimal use case. There are many areas that we didn’t address yet, including how to support use cases like yours. You are more than welcome to join our regular meetings (Monday 9am EDT) and contribute to our discussions.

Regards,

Ifat Afek

Software Engineer

CloudBand Business Unit

Alcatel-Lucent

16 Atir Yeda St. Kfar Saba 44643, Israel

T: +972 9 793 3085

F: +972 9 793 3700

ifat.afek@alcatel-lucent.com

From: tosca@lists.oasis-open.org [mailto:tosca@lists.oasis-open.org] On Behalf Of Luca Gioppo
Sent: Monday, June 08, 2015 1:26 PM
To: TOSCA
Subject: [tosca] monitoring

I'm interested in the monitoring topic.

My concern is that on some docs I've red it seems that having the monitoring agents and server system is considered a precondition.

Have I understood well?

If yes I do not agree since we are installing the application from scratch on new VMs so nothing is there unless TOSCA orchestrator places it (unless we use some precooked images, but doing so we are shortcutting the TOSCA standard since there is something that falls outside the application description).

Also tipycally (I'm using Zabbix) monitoring is something that gets achieved by instrumenting the agent and the monitoring server and in some cases adding some instructions in the software component (activating Apache status for example).

I think is more than declaring what we are monitoring but we need to define how we configure the specific agent/server that could require opening ports and so on.

At which level the standard want to set itself?

I think that the standard can give for granted the existance of a server for monitoring somewhere (so we do not bother where is the zabbix server) but we need to add to the topology an "agent node type" related with the host container to allow us to install the agent and configure it (in the end is just another software we need to install) and than we need the relations to allow it to look at the measurements we need (in the zabbix example we have the template concept that is a group of measurements and more) so a relation towards Apache where we will need a proper endpoint and will need to examine the info, or a relation to the host to get cpu usage.

Luca

tosca message