NSP Flow Collector and Flow Collector Controller fault tolerance

Description

You can deploy redundant NSP Flow Collector Controllers in separate data centers to manage local Flow Collectors, or a common set of Flow Collectors that are deployed in logical proximity to the managed NEs, and not assigned to a data center. The latter configuration essentially eliminates switchover latency and any resulting data loss in the event of a Flow Collector Controller failure. The configuration also has no dependency on which Flow Collector Controller currently has the primary role.

The NSP Flow Collector Controllers continuously monitor the active nspOS; if the connection is lost or re-established after a connection loss, the Flow Collector Controller primary and standby roles are re-evaluated. In a DR deployment, redundancy is ensured only when the two Flow Collector Controllers are deployed in separate data centers, in order that the failure of one data center does not cause both Flow Collector Controllers to be unreachable.

Redundant NSP Flow Collectors in a data center can collect statistics from the same set of NEs to provide data-collection fault tolerance. The statistics that are stored in a database are de-duplicated beforehand; only one set is stored. If statistics are stored in files, duplicate files are created.

In an HA or HA+DR deployment that includes the NFM-P, the primary Flow Collector Controller connects only to the primary NFM-P. An internal mechanism ensures that the primary Flow Collector Controller in a DR deployment is always in the same data center as the primary NFM-P.

Note: An NSP switchover or failover disrupts the NSP API communication with one or more Flow Collectors in SYS mode. As a result, Flow Collector log entries like the following are recorded at a high frequency:

Connection to node node (IP_address) could not be established. Broker may not be available

The issue is not service-affecting. To stop the recording of the log entries, you must restart the Flow Collector.

Remote statistics transfer

An NSP Flow Collector can transfer collected statistics files to redundant remote servers. For greater fault tolerance, redundant Flow Collectors can collect statistics from the same set of NEs and transfer the files to redundant destination servers. Such a configuration ensures that the statistics collection and transfer continue uninterrupted in the event that an NSP Flow Collector and a transfer destination are each unreachable.