Once a device is modeled and actively monitored via SNMP no longer requires ICMP! Why is this? If we lose polling to the device then it seems to
me a critical Device has stopped responding to polls and external request type alarm will be raised instead of management agent lost - an unnecessary criticality.
Release : 10.4.1
Component : Spectrum Core / SpectroSERVER
The SpectroSERVER will only poll SNMP based models using SNMP, it will not poll using ICMP unless snmp
stops responding to determine if the device is still up. If the device does not respond to SNMP but does
respond to ICMP a 'Management Agent Lost' alarm is raised. If the devices not respond to SNMP and ICMP
then a 'Device Has Stopped Responding to Polls' Critical Alarm is raised.
For ICMP based models (Pinables), the SpectroSERVER will only poll via ICMP and raise a 'Device
Has Stopped Responging to Polls' Critical alarm (so long as the models value_when_red/0x1000e is set to 7 and not 0)>
For Discovery
'Model By IP' / 'Model By Host'
will discover using SNMP and only use ICMP should the SNMP requests timeout.
Discovery Console
Has an option to use ICMP first, then poll with SNMP. This helps when large ranges are being discovered and there is
a list of community strings to try.
Under 'Advanced Options'
Fault Isolation
The SpectroSERVER will use ICMP when checking connected neighbors after a device has been found to have stopped responding to all polls.