Hi all,
in case this is a single device only this may point indeed to a device problem. What you can do - ahead of a sniffer running besides the SpectroSERVER filtered for the device IP - you can verify if that device is maybe busy or stressed.
A device could be stressed by SNMP request - and to verify lower that workload check for:
- attribute "Throttling" / 0x11f79 - Boolean - and set to "yes" (default is "no") - along with
- attribute "Throttle_Count" / 0x11f39 - Integer - and set this to "5" (default is 10)
Once this is done CA Spectrum will throttle the SNMP-requests when send to the device. In addition to this - go to the Device' Component Details Information tab - find Subview "CA Spectrum Modeling Information" and set here DCM Timeout to 5000 msec (default 3000ms).
Now verify if you still see the period Management Agent lost alarms. In case you wont - then reduce DCM timeout back to i.e. 3000 - or 2000 ms. Still having no Management Agent lost then indicates truly this is a device SNMP request stress issue. Smal / SOHO devices may show that symptom - or just overloaded devices. The timeout is related to the "network" between SpectroSERVER and the device - the throttling addresses a SNMP-stress condition. So you have 2 sets of parms (network connectivty and "delay" - and the SNMP related "workload at device level") you can tune.
In case that tuning wont affect the symptom - then we have to look at the sniffer data to see if the device may return a negative response - or just "timeout" ..
Kind regards, Joerg
(It would be good to know which devices this are - that may allow CA to check into the knowledgebase.)