Synchronization for multiple data sources

If you run a discovery that includes collector and SNMP data sources, or multiple collectors, you must synchronize the "Interrogating Devices" phase (phase 1) of the discovery across all of the data sources. Synchronization of phase 1 reduces the likelihood of excessive discovery cycling and ensures that the maximum amount of data is available during topology building. Synchronization capability is provided by the m_WaitForManagedProcs configuration setting. m_WaitForManagedProcs is set to 0 (off) by default, which is more suited to SNMP-only discoveries.

During phase 1 of an IP discovery, discovery waits by default up to 90 seconds between the discovery of one device and the next. If no further devices are found after 90 seconds, then the discovery moves into phase 2. This time delay is controlled by the m_NothngFndPeriod field in the disco.config discovery database table, and is ideal for SNMP discoveries in which entities do not all respond at the same time.

However, the behavior of collectors is different than SNMP. Collectors return all of their devices in one go, potentially after a comparatively long initial delay, during which the EMS is queried. Using only m_NothngFndPeriod when multiple collectors or collector and SNMP discoveries are run under the same domain can lead to undesired multiple discovery cycles. This effect happens because the interrogation phase delay between each collector can be far in excess of 90 seconds. This delay causes the discovery to continue to the next phase if the SNMP ping phase completed.