Help us improve your experience.

Let us know what you think.

Do you have time for a two-minute survey?

Navigation
Guide That Contains This Content
[+] Expand All
[-] Collapse All

    High Availability for Analytics

    Contrail supports multiple instances of analytics for high availability and load balancing.

    Contrail analytics provides two broad areas of functionality:

    • contrail-collector —Receives status, logs, and flow information from all Contrail processing elements (for example, generators) and records them.

      Every generator is connected to one of the contrail-collector instances at any given time. If an instance fails (or is shut down), all the generators that are connected to it are automatically moved to another functioning instance, typically in a few seconds or less. Some messages may be lost during this movement. UVEs are resilient to message loss, so the state shown in a UVE is kept consistent to the state in the generator.

    • contrail-opserver —Provides an external API to report UVEs and to query logs and flows.

      Each analytics component exposes a northbound REST API represented by the contrail-opserver service (port 8081) so that the failure of one analytics component or one contrail-opserver service should not impact the operation of other instances.

      These are the ways to manage connectivity to the contrail-opserver endpoints:

      • Periodically poll the contrail-opserver service on a set of analytics nodes to determine the list of functioning endpoints, then make API requests from one or more of the functioning endpoints.
      • Subscribe to the Contrail Discovery Service to get a list of functioning endpoints. If there are any failures, it can take 5-30 minutes for the Contrail Discovery Service to send an update.

        The Contrail user interface makes use of the same northbound REST API to present dashboards, and reacts to any contrail-opserver high availability event automatically, using the Contrail Discovery Service.

    Modified: 2015-09-02