MONitoring as a Service for OpenStack

In early May, we announced the publication of a blueprint proposing the integration of a monitoring service to the OpenStack cloud platform. Called Monitoring as a Service (MONaaS), this project was presented for the first time on May 21 at the Solutions Libres & Open Source Fair in Paris, France, by Thibault Cohen and myself.

Here is the slide deck and we will be presenting it again tomorrow (June 4) at the third meeting of the OpenStack Montreal community.

Today, in fact, there is no supervision solution in OpenStack except through the Ceilometer project. This OpenStack component is designed to collect measures of the various components of the platform – originally for billing purposes. It meets the  operators’ needs of measuring the cloud infrastructure, but it is not a monitoring solution for users of the infrastructure and their applications.

Since ceilometer does not allow monitoring at the application level, we propose to add a component of Monitoring as a Service (MONaaS) to OpenStack.

• It would be a RESTful, fault-tolerant, self-supervised and highly available service.
• This service would allow users to create different audit operations that would be processed out on a regular basis by the monitoring service.
• An optional agent collecting application data that would otherwise not be visible from the hypervisor would be installed in the instances.
• The monitoring service could also carry out availability checks by directly testing services on their TCP / UDP standards ports (SMTP, HTTP, etc…) and even perform web scenarios.

Of course, this contribution is fully open and we are currently discussing with the technical community to better integrate our project in the OpenStack ecosystem. Pending integration of Monitoring as a Service in Openstack, Savoir-faire Linux released a Nagios plugin that allows metrics monitoring in Ceilometer.

▸ This post has been prepared and co-written by Alexandre Viau and Thibault Cohen

Video of the presentation (UPDATED)