2.16. Telemetry

This section outlines the top new features and changes for the Telemetry service.
OpenStack Telemetry Metrics (gnocchi) at Scale
Telemetry used MongoDB and Telemetry API to store metrics, while the performance was acceptable when it came to storing the metrics, the usage is limited as you cannot retrieve and exploit the stored information.
The OpenStack Telemetry Metrics (gnocchi) service uses a new distributed selective acknowledgements (SACKs) mechanism and scheduling algorithm for the gnocchi-metricd daemon improving the performance at a larger scale. The default settings are enhanced to adapt to cloud deployment of larger sizes.
Intel Cache Monitoring Technology (CMT)
Cache Monitoring Technology (CMT) allows you to monitor cache-related statistics on an Intel platform. Telemetry now supports CMT reporting using the collectd daemon.
This release adds a new meter to collect the L3 cache usage statistics for each virtual machine. You can enable the cmt plugin with LibvirtEnabledPerfEvents parameter in nova-libvirt.yaml file.
Containerization of Telemetry Services
With this release, Red Hat OpenStack Platform can create a cloud that uses containers to host its services. Each service is isolated within its own container on the host node. Each container connects to and shares the host’s own network. As a result, the host node exposes the API ports of each service on its own network. Telemetry service can now be hosted on containers. This makes upgrades easy.
OpenStack Telemetry Event Storage (panko) Deprecation
The OpenStack Telemetry Event Storage service is officially now deprecated. Support for panko will be limited to usage from Red Hat Cloudforms only. Red Hat does not recommend using panko outside of the Red Hat Cloudforms use-case. You can use the following options instead of using panko:
  • Poll the OpenStack Telemetry Metrics (gnocchi) service instead of polling panko. This gives you access to the resource history.
  • Use the OpenStack Telemetry Alarming (aodh) service to trigger alarms when an event arises. You can use OpenStack Messaging Service (zaqar) to store alarms in a queue if an application cannot be reached directly by the OpenStack Telemetry Alarming (aodh) service.
Telemetry API and ceilometer-collector Deprecation
The Telemetry API service is now deprecated. It is replaced by the OpenStack Telemetry Metrics (gnocchi) service, and the OpenStack Telemetry Alarming (aodh) service APIs. You should begin to switch to the Telemetry API service instead. In Red Hat OpenStack Platform 12, the Telemetry API is disabled by default, with the option to enable it only if required.
The ceilometer-collector service is deprecated. You can now use the ceilometer-notification-agent daemon because the Telemetry polling agent sends the messages from the sample file to the ceilometer-notification-agent daemon.
NOTE: Ceilometer as a whole is not deprecated, just the Telemetry API service and the ceilometer-collector service.