Spectrum will send a SNMP get-request or SNMP get-next-request to a device at each poll interval (configured via the "Polling_Interval" attribute (0x10071). Each request will have a unique request-id. Spectrum waits for a response to this (the time that Spectrum waits is configured on the model in the "TimeOut" attribute (0x110c4). By default the value is 3000 ms, or 3 seconds). If no response is received (with the same request-id) within that timeframe, then Spectrum will determine if another SNMP get-request is needed (Spectrum determines this by the value specified in the TryCount attribute (0x110c5). By default this value is 3). If another request should be sent, Spectrum will send the request with a NEW unique request-id. And the process is repeated until the Timeout and TryCount values have been reached. Once they have been reached, the Management Agent Lost alarm is generated.
For example, let's say for a device that the poll interval is 5 minutes, the DCM Timeout and DCM Trycount values are 3000ms and 3.
At the 5 minute mark, Spectrum sends an SNMP get-request with a timestamp of 9:43:14 to the device and that SNMP has a request-id of 18846920..
No response is received within 3 seconds, so another SNMP get-request is sent to the device with request-id of 18847047 and a timestamp of 9:43:17.
The device sends a get-response to Spectrum with a request-id of 18846920 and it has a timestamp of 9:43:20. This is 6 seconds after the SNMP get-request was sent with the same request-id. Since this is over the 3000ms mark, Spectrum does not consider this to be acceptable contact.
Spectrum sends another SNMP get-request at 9:43:20 with a request-id of 18847076.
The device sends a get-response at 9:43:23 with a request-id of 18847047. This is 5 seconds after the SNMP get-request was sent with the same request-id so again, Spectrum does not consider this to be acceptable contact.
The device sends a get-response to Spectrum with a request-id of 18847076 and it has a timestamp of 9:43:26. This is 6 seconds after the SNMP get-request was sent with the same request-id so again, Spectrum does not consider this to be acceptable contact and since the TryCount value has now been reached (3) Spectrum generates a Management Agent Lost Alarm.
What process does Spectrum use to generate a Management Agent Lost alarm?
How does Spectrum do to determine that SNMP communication has been lost with a device?
Management Agent Lost alarms are generating even though it looks like contact to the device is established.
(Legacy KB ID CNC TS33637 )