We have observed that the CM_NIMBUS_PROBE and CM_NIMBUS_PACKAGE database tables are not being updated in a timely manner. For example, when we stop a probe and check the status for a specific robot in the database, the records are not updated after 30 minutes in our production environment. We rely extensively on these database queries to check the status of probes and packages. We have noticed that the probe_discovery queue occasionally becomes stuck; while it shows as green and connected, it does not drain until the queue is restarted.
DX UIM 23.4.8 (CU8) and earlier
Performance degradation in database table updates and probe_discovery queue processing is typically caused by excessive re-correlation events triggered by frequent device attribute changes (such as multiple MAC or IP addresses) and memory pressure on the discovery_server probe.
Apply Discovery Server Patch
Deploy the attacheddiscovery_server build (discovery_server_23.4.6.1.zip) which includes optimizations for handling re-correlation events. This build handles recreation events more efficiently, ensuring that incoming devices with high attribute volatility do not impact the overall performance of the probe.
Optimize Database Connection Settings If delays in database updates persist, adjust the connection pool settings in the discovery_server raw configuration to improve processing throughput:
discovery_server probe.setup section.db_connection_pool_max_idle to 50.db_connection_pool_size to 50.IMPORTANT NOTE
This fix is build for DX UIM 23.4CU6.
DX UIM 23.CU9 and later will include this enhancements. If a fix is needed for CU7 or CU8, Contact Broadcom support