Track node CPU and memory utilization
In order to track the ~"group::memory" NSM more accurately, we need to not just track the total number of cores and total available memory per node, but also to what extent they're used. This will help us understand if self-managed customers are over- or underprovisioned in terms of hardware an w.r.t. our reference architectures.
See also gitlab#230898 (closed)
I think we can largely base this on the existing node_exporter
metrics, but it did not have a utilization ratio for memory, so I added that to node.rules
.
I then created two usage ping specific aggregations that we will query back from the client.
Edited by Matthias Käppler