- Docs
- How-to guides
- Identify idle servers
How-to guides
Identify idle servers
Spot servers running flat, decide whether to keep, downsize, or stop them.
Last updated: 2026-05-21
On this page
An idle server is one that's billed continuously but doing very little — low CPU, low network, no business reason to keep running. They're the most common form of cloud waste on Hetzner because stopped servers still cost money (the volumes and IPs stay attached) and forgotten servers are easy to miss.
CloudTally surfaces idle candidates in two places: a count on the dashboard, and a per-server view in Resources. What you can do from each depends on your plan.
What "idle" means in CloudTally
CloudTally calls a server idle when CPU utilisation has stayed below a threshold for an extended period — long enough that it's not just a quiet night. The exact rule is documented in Waste detection rules (coming soon), but the short version: sustained low CPU over multiple days.
The signal is conservative. CloudTally would rather miss a true idle server than wake you up about a quiet web server with bursty traffic. If you suspect a server is idle but it's not flagged, check the Inspect drawer in the Cost Explorer for its CPU history.
On Free: see the count, act on the obvious
Free users get the idle-server count on the dashboard. It looks like a small badge with a number and a "review" link. Clicking opens the Resources page, sorted by monthly cost.
What you can do from here:
- Sort by Monthly cost (descending) to find the most expensive servers first.
- Click into each server to see its details, including the project it belongs to and when it was created.
- Stop or delete servers in the Hetzner Cloud Console. (CloudTally never touches your account — it's strictly read-only.)
What you can't do from the Free view:
- See which specific servers CloudTally classified as idle (the count tells you there are some, not which).
- See the projected monthly saving if you stopped them.
- Filter directly to idle candidates without scrolling through the full inventory.
For two or three obviously idle servers — names like temp-test-rebuild or old-staging — the count plus your own judgement is enough. For ten or more, you'll spend more time hunting than the trial is worth.
On Pro: full drilldown
Pro unlocks the waste drilldown view. The dashboard idle badge becomes clickable into a dedicated list with:
- The exact set of servers CloudTally identified as idle, with their CPU utilisation history.
- A projected monthly saving per server, computed from the resource's current cost and assumed full reclamation.
- Bulk actions: select multiple servers and copy a list ready to paste into the Hetzner Console for action.
This is the most common driver of upgrade decisions: once you've seen the drilldown once and acted on it, the saving usually pays for several months of Pro.
A short audit you can do today
Whether you're on Free or trial-Pro, this five-minute audit usually pays back:
- Open the dashboard's idle badge.
- For each server you can identify as idle, ask: is anything still running on this?
sshin and check the process list and last login. - If nothing's there, stop the server in Hetzner. You'll still pay for the volumes and the IP, but the server-hours stop accumulating.
- After two weeks, if no one has asked you to bring it back, delete it.
Stopping first, then deleting after a delay, is the safest pattern. It catches the case where someone needed that server but hadn't told you.
Related how-tos
- Handle unattached volumes — volumes that survived a deleted server.
- Review rightsizing recommendations — for servers that are doing work, just too much server for the load.
- Investigate a cost spike — when waste is part of a broader cost increase.