I know this has been discussed on numerous threads and short of missing something in those threads, I'm still left w/ a question.
1. using agent-based polling
2. not monitoring buffers, cached or shared memory
I know about this:
Help! Linux ate my RAM!
The node is monitored via the agent. I'm dealing w/ an onslaught of critical alerts repeatedly. The nodes are fine but the NOC team is obliged to respond to each.
Many of my vm's are reporting critical physical memory consumption, but in reality, they are fine.
I have the following monitored
This is what memory shows me on the server itself
Ideas?