Standard setup. Node Exporter and cAdvisor running on the servers, feeding metrics to the monitoring server over Wireguard.
Nothing exotic. I wasn’t chasing a specific failure mode when I set it up. I just wanted numbers to look at.
The alert
Disk usage on the Samba share on the NAS kept tripping the threshold. Almost full, every time.
First reaction: something’s wrong. Logs filling up? A backup job stuck in a loop? I went looking for a problem.
There wasn’t one.
What was actually happening
Time Machine. It grabs whatever space it’s assigned on the Samba share and just… uses it. All of it, basically all the time.
That’s how Time Machine works. It expands to fill the quota, then manages its own history within that space. From macOS’s side, that’s completely normal.
From Node Exporter’s side, it looks identical to “disk about to fill up and break something.”
What I Learned
A generic disk-usage alert doesn’t know the difference between “about to run out of space” and “working exactly as intended.” It just sees a number crossing a line.
I could tune the threshold higher. But that just moves the false alarm, it doesn’t remove it. The real fix is knowing which shares are supposed to run near-full and excluding them from that specific alert, or giving them their own threshold entirely.
I hadn’t thought about that distinction until the alert forced me to.
Conclusion
Nothing was broken. The monitoring did exactly what I asked it to do — it just didn’t know what “full” was supposed to mean for that share.
Alerts are only as smart as the assumptions you gave them.