Instance Discovery vs. Polling issue

Started by Tursiops, April 04, 2017, 03:37:16 AM

Previous topic - Next topic

Tursiops

Hi,

I seem to be having an issue where instances just disappear.
The reason appears to be that if more than one polling method is enabled, unless all of them fail, the node is not considered to be down.

If the node isn't down, it'll happily run Instance Discovery.
Of course if the agent or SNMP do not respond (but the other does, or ICMP may be working and part of polling), then instance discovery will fail to discover any instances.
That's when it seems to wipe any existing instances.

Has anyone else encountered this?
If that's not just some oddity at our end, maybe instance discovery should only run for agent related items when the agent is considered to be up, same for SNMP. If the agent or SNMP do not respond when instance discovery starts, the results should probably be discarded, to avoid losing all history on existing instances?

Cheers,
Tursiops