Data collection¶

Compute agent¶

This agent is responsible for collecting resource usage data of VM instances on individual Compute nodes within an OpenStack deployment. This mechanism requires a closer interaction with the hypervisor, therefore a separate agent type fulfills the collection of the related meters, which is placed on the host machines to retrieve this information locally.

A Compute agent instance has to be installed on each and every compute node, installation instructions can be found in the Install the Compute agent for Telemetry section in the Installation Tutorials and Guides.

Just like the central agent, this component also does not need a direct database connection. The samples are sent via AMQP to the notification agent.

The list of supported hypervisors can be found in Supported hypervisors. The Compute agent uses the API of the hypervisor installed on the Compute hosts. Therefore, the supported meters may be different in case of each virtualization back end, as each inspection tool provides a different set of meters.

The list of collected meters can be found in OpenStack Compute. The support column provides the information about which meter is available for each hypervisor supported by the Telemetry service.

Middleware for the OpenStack Object Storage service¶

A subset of Object Store statistics requires additional middleware to be installed behind the proxy of Object Store. This additional component emits notifications containing data-flow-oriented meters, namely the storage.objects.(incoming|outgoing).bytes values. The list of these meters are listed in OpenStack Object Storage, marked with notification as origin.

The instructions on how to install this middleware can be found in Configure the Object Storage service for Telemetry section in the Installation Tutorials and Guides.

Telemetry middleware¶

Telemetry provides HTTP request and API endpoint counting capability in OpenStack. This is achieved by storing a sample for each event marked as audit.http.request, audit.http.response, http.request or http.response.

It is recommended that these notifications be consumed as events rather than samples to better index the appropriate values and avoid massive load on the Metering database. If preferred, Telemetry can consume these events as samples if the services are configured to emit http.* notifications.

Central agent¶

This agent is responsible for polling public REST APIs to retrieve additional information on OpenStack resources not already surfaced via notifications, and also for polling hardware resources over SNMP.

The following services can be polled with this agent:

• OpenStack Networking
• OpenStack Object Storage
• OpenStack Block Storage
• Hardware resources via SNMP
• Energy consumption meters via Kwapi framework

To install and configure this service use the Add the Telemetry service section in the Installation Tutorials and Guides.

The central agent does not need direct database connection. The samples collected by this agent are sent via AMQP to the notification agent to be processed.

IPMI agent¶

This agent is responsible for collecting IPMI sensor data and Intel Node Manager data on individual Compute nodes within an OpenStack deployment. This agent requires an IPMI capable node with the ipmitool utility installed, which is commonly used for IPMI control on various Linux distributions.

An IPMI agent instance could be installed on each and every Compute node with IPMI support, except when the node is managed by the Bare metal service and the conductor.send_sensor_data option is set to true in the Bare metal service. It is no harm to install this agent on a Compute node without IPMI or Intel Node Manager support, as the agent checks for the hardware and if none is available, returns empty data. It is suggested that you install the IPMI agent only on an IPMI capable node for performance reasons.

Just like the central agent, this component also does not need direct database access. The samples are sent via AMQP to the notification agent.

The list of collected meters can be found in Bare metal service.

Notification agent HA deployment¶

In the Kilo release, workload partitioning support was added to the notification agent. This is particularly useful as the pipeline processing is handled exclusively by the notification agent now which may result in a larger amount of load.

To enable workload partitioning by notification agent, the backend_url option must be set in the ceilometer.conf configuration file. Additionally, workload_partitioning should be enabled in the Notification section in the OpenStack Configuration Reference.

Polling agent HA deployment¶

The availability check of the instances is provided by heartbeat messages. When the connection with an instance is lost, the workload will be reassigned within the remained instances in the next polling cycle.

For backward compatibility and supporting existing deployments, the central agent configuration also supports using different configuration files for groups of service instances of this type that are running in parallel. For enabling this configuration set a value for the partitioning_group_prefix option in the polling section in the OpenStack Configuration Reference.

To enable the Compute agent to run multiple instances simultaneously with workload partitioning, the workload_partitioning option has to be set to True under the Compute section in the ceilometer.conf configuration file.

Meter definitions¶

The Telemetry service collects a subset of the meters by filtering notifications emitted by other OpenStack services. Starting with the Liberty release, you can find the meter definitions in a separate configuration file, called ceilometer/meter/data/meter.yaml. This enables operators/administrators to add new meters to Telemetry project by updating the meter.yaml file without any need for additional code changes.

A standard meter definition looks like:

---
metric:
  - name: 'meter name'
    event_type: 'event name'
    type: 'type of meter eg: gauge, cumulative or delta'
    unit: 'name of unit eg: MB'
    volume: 'path to a measurable value eg: $.payload.size'
    resource_id: 'path to resource id eg: $.payload.id'
    project_id: 'path to project id eg: $.payload.owner'

The definition above shows a simple meter definition with some fields, from which name, event_type, type, unit, and volume are required. If there is a match on the event type, samples are generated for the meter.

If you take a look at the meter.yaml file, it contains the sample definitions for all the meters that Telemetry is collecting from notifications. The value of each field is specified by using JSON path in order to find the right value from the notification message. In order to be able to specify the right field you need to be aware of the format of the consumed notification. The values that need to be searched in the notification message are set with a JSON path starting with $. For instance, if you need the size information from the payload you can define it like $.payload.size.

A notification message may contain multiple meters. You can use * in the meter definition to capture all the meters and generate samples respectively. You can use wild cards as shown in the following example:

---
metric:
  - name: $.payload.measurements.[*].metric.[*].name
    event_type: 'event_name.*'
    type: 'delta'
    unit: $.payload.measurements.[*].metric.[*].unit
    volume: payload.measurements.[*].result
    resource_id: $.payload.target
    user_id: $.payload.initiator.id
    project_id: $.payload.initiator.project_id

In the above example, the name field is a JSON path with matching a list of meter names defined in the notification message.

You can even use complex operations on JSON paths. In the following example, volume and resource_id fields perform an arithmetic and string concatenation:

---
metric:
- name: 'compute.node.cpu.idle.percent'
  event_type: 'compute.metrics.update'
  type: 'gauge'
  unit: 'percent'
  volume: payload.metrics[?(@.name='cpu.idle.percent')].value * 100
  resource_id: $.payload.host + "_" + $.payload.nodename

You can use the timedelta plug-in to evaluate the difference in seconds between two datetime fields from one notification.

---
metric:
- name: 'compute.instance.booting.time'
  event_type: 'compute.instance.create.end'
 type: 'gauge'
 unit: 'sec'
 volume:
   fields: [$.payload.created_at, $.payload.launched_at]
   plugin: 'timedelta'
 project_id: $.payload.tenant_id
 resource_id: $.payload.instance_id

You will find some existence meters in the meter.yaml. These meters have a volume as 1 and are at the bottom of the yaml file with a note suggesting that these will be removed in Mitaka release.

For example, the meter definition for existence meters is as follows:

---
metric:
  - name: 'meter name'
    type: 'delta'
    unit: 'volume'
    volume: 1
    event_type:
        - 'event type'
    resource_id: $.payload.volume_id
    user_id: $.payload.user_id
    project_id: $.payload.tenant_id

These meters are not loaded by default. To load these meters, flip the disable_non_metric_meters option in the ceilometer.conf file.

Database dispatcher¶

When the database dispatcher is configured as data store, you have the option to set a time_to_live option (ttl) for samples. By default the time to live value for samples is set to -1, which means that they are kept in the database forever.

The time to live value is specified in seconds. Each sample has a time stamp, and the ttl value indicates that a sample will be deleted from the database when the number of seconds has elapsed since that sample reading was stamped. For example, if the time to live is set to 600, all samples older than 600 seconds will be purged from the database.

Certain databases support native TTL expiration. In cases where this is not possible, a command-line script, which you can use for this purpose is ceilometer-expirer. You can run it in a cron job, which helps to keep your database in a consistent state.

The level of support differs in case of the configured back end:

HTTP dispatcher¶

The Telemetry service supports sending samples to an external HTTP target. The samples are sent without any modification. To set this option as the collector’s target, the dispatcher has to be changed to http in the ceilometer.conf configuration file. For the list of options that you need to set, see the see the dispatcher_http section in the OpenStack Configuration Reference.

File dispatcher¶

You can store samples in a file by setting the dispatcher option in the ceilometer.conf file. For the list of configuration options, see the dispatcher_file section in the OpenStack Configuration Reference.