As you can see in the graph, like 65% of memory usage is ARC. This is fine, but the web UI reports this as actual used memory, leading to incorrect usage stats. Glances reports it correctly.
As you can see in the graph, like 65% of memory usage is ARC. This is fine, but the web UI reports this as actual used memory, leading to incorrect usage stats.
Basically NOT used Ram is wasted Ram. And ARC is clever enough to shrink automatically if the system needs more Ram. Unfortunately this shrinking is slow - a hefty request of more Ram may fail and the OOM handler might do its job...
My personal rule: you cannot over-commit Ram as you do with CPU and (possibly) storage. At least not more than a few percent. Depending on the actual use case of course...
Looks totally fine to me. You're using roughly 60GB without L2ARC, rest is L2ARC usage (~ 170GB) + some remaining free GB (~ 20). I see absolutely nothing wrong here.
The graphs have different color. If you hover over these graphs it will tell you that if you select the upper line thats your RAM (and also without L2ARC), if you move a bit down and select the second line it (turkise) it says L2ARC
E.g.
Because the RAM is activly beeing used (the reason what it is beeing used for doesnt matter), the graph shows it as used, split into each different form of usage.