twest Yeah, this is what we need to know.
Sounds like the common misconception of Linux memory usage. Most OS monitoring solutions either can't do Linux memory monitoring properly, or are badly configured as it near always shows used + cached. I still get DevOps engineers running to me that a production server is about to fall over because it's at 90%+ RAM... Well, no, that's not the case - no matter how many times I've explained the case.
For example: My control panel server has 8GB of memory assigned to it, Grafana is showing it at 81% memory usage but if I dig into the graphs I can see only 2.5GB is actually used and 4GB is sitting in the cache. The cache can take as much as it wants/needs as it is immediately released as soon as a process needs it.
