Originally Posted by NatalieMojoHost

The deal with Linux is: it will try to use all of the RAM in the system for speeding up the filesystem and other things. That's what the buf/cache is.

The shortest answer is - you have 79GB available and 43GB hard in-use, but Linux is taking another 65GB and using it to slightly speed up your system in other ways.

If another process needs that memory, say MySQL or Elasticsearch, it's able to pull it out of that buffer/cache pool and away from the filesystem. But it will pull it from the 16GB free first.

Hope this helps clear things up for you.

Originally Posted by wankawonk

solid answer

I would recommend just using top, add up the "free" and "buf/cache" columns and that's how much you have "free". though you should always leave some buf/cache to speed up any potential swapping.

Important elasticsearch detail: If your elasticsearch shard size exceeds the amount of heap space you allocate to it, elasticsearch will get a big speed boost from having memory available for buf/cache. So be careful to monitor that. don't just assign all your buf/cache to (for example) redis because it might severely impact your elasticsearch performance.