Every node's agent samples its servers every 5 minutes. From those samples the panel draws the usage charts on each server's Usage tab, counts each server's traffic per calendar month, and applies the plan's rules when a server passes its allowance. Administrators and resellers see every server's traffic on one page, Servers › Traffic.
The panel shows this above each server's usage section, and it is the rule throughout: bandwidth usage and stats are updated every 5 minutes, the meter resets at 00:00 UTC on the 1st of each month, and all stats are in UTC.
How usage is measured#
- The agent samples every server every 5 minutes, aligned to the clock, with one libvirt call for all servers: network bytes in and out (as seen from the server: in is received by the server), CPU time, memory, disk reads and writes, and per-disk throughput and IOPS. It keeps the last 48 hours on the node, so nothing is lost while the panel is away.
- The panel collects the samples from each node every 5 minutes, through the
nodesqueue workers. A node that was away is caught up from its 48 hours. - Samples are kept as 5-minute points (8 days by default), hourly points (45 days) and daily points (400 days). The retention can be raised under System › Settings › Logs (the three Usage rows); the minimums cannot be lowered, because the charts need that much.
- Memory use needs the guest to report it (the balloon driver or guest agent). Without it the RAM chart shows the allocation only and says "The server does not report memory use; install the guest agent".
Note: If charts stop moving for every server of a node, the samples are not being collected. Check that the node is online and that the
nodesworkers run: see Workers and queues.
Usage charts#
On a server's Usage tab:
- Bandwidth (in and out), CPU (% of the server's vCPUs), RAM (used of allocated), HDD read and write, and HDD space used of the disk size.
- Spans: 1 hour, 8 hours and 1 day (5-minute points), 1 week and 1 month (hourly), 1 year (daily). Times are UTC.
- Disks: for each disk of the server, a Throughput chart and an IOPS chart (read and write), with the disk's I/O limits drawn as dashed lines when it has any. They follow the span chosen above. Per-disk figures need a recent agent; older agents show "No per-disk figures yet".
Charts are also available through the API: GET /api/v1/servers/{id}/usage?span= and GET /api/v1/servers/{id}/disk-usage?span= (scope usage:read).
Traffic this month#
The Usage tab shows Traffic this month: in plus out (both count) against the server's allowance, as a meter, with the reset date ("Resets on 1 Dec at 00:00 UTC").
- An allowance of N GB is N × 1024³ bytes, like every size in the panel.
- A server given to another account mid-month: what it used before stays counted for its previous owner and pool; only what it uses afterwards counts for the new one. The server's own meter counts the whole month.
- A deleted server's traffic, and a deleted account's, keep counting for their reseller's pool until the month resets. Up to 5 minutes of traffic just before a server is deleted are not counted.
- Reseller pools count measured traffic against the reseller's traffic limit. A pool that has used it all refuses new servers until the 1st.
Past the allowance#
Each plan says what happens under When a server passes its traffic:
| Option | What happens |
|---|---|
| Tell its owner, nothing more (default) | Notices only. |
| Limit its speed until the month resets | Both directions are limited to Limited to (Mbit/s) (10 Mbit/s unless set). A throttle never raises a limit that is already lower. |
| Suspend it until the month resets | The server is suspended with the reason "Traffic allowance used up this month". |
At 80 % and 100 % the owner and their reseller get a notice (once each per server and month) on their dashboard, in the server's Usage tab and by mail. The server.traffic_threshold webhook fires at both points. Administrators find the notices in the audit log (vm.traffic_notice).
A throttle or a traffic suspension ends:
- when the month resets (checked every 5 minutes);
- when an administrator clicks Lift the limit or Lift the suspension on the Usage tab (or Lift the limit on the Traffic page). The server is then not throttled or suspended again that month;
- when the cause goes: the plan no longer acts past the traffic, or the server's allowance grows.
A suspension that staff placed is never lifted by any of these; the traffic suspension waits beside it. See Servers for suspensions in general.
Speed limits#
- Plans have Speed in and Speed out in Mbit/s (empty for no limit; 1 to 100000). See Plans.
- A server follows its plan's speed limits and overage setting. Saving a plan with other values copies them to its servers and sends the new limits to each running or stopped server.
- Administrators change one server's traffic allowance, overage action, throttle speed and speed limits with Change limits on its Usage tab. A later change of the plan's speed limits replaces them. Resellers cannot change one server's limits.
- If a node reports other limits than a server should have (after rescue, or a change while the server was being built), the panel sends them again when the server has no other task. The Usage tab says when the last change did not go through.
The Traffic page#
Servers › Traffic lists every server's traffic for one calendar month. Administrators see every server; resellers see their own servers and their customers'. Customers do not have this page; their figures are on each server's Usage tab and on Usage & limits.
- Month (UTC): this month by default, or any of the last 12.
- Columns: the server (linking to its Usage tab) and its owner; node and location (administrators only); plan; in; out; Counted against the allowance with a meter; speed limits, and whether the server is throttled or suspended for its traffic.
- Totals above the table for the filtered servers: count, in, out, counted, and how many are near, over and throttled.
- Filters: search (hostname, IP address or owner), state, plan, owner, and for administrators node and location.
| State filter | Servers |
|---|---|
| Near the allowance (80% to 99%) | 80 % to under 100 % of their allowance |
| Over the allowance | 100 % or more |
| Throttled or suspended for traffic | Held for their traffic now |
| Speed limited | Any speed limit of their own, or a throttle |
| No allowance | No traffic allowance set |
- Sorting: click a column header. The default is counted traffic, largest first, so the top users are at the top.
- Speed limit (administrators): opens a short form with Speed in (Mbit/s) and Speed out (Mbit/s) (empty for no limit), and Lift the limit when the server is held for its traffic. Click Save speed limit. It is the same change as on the Usage tab, audited as
vm.limits_changed. - Export CSV downloads every row the filters select, in the current sort, with bytes as whole numbers and times in UTC. Resellers' exports have no node and location columns.
Through the API (administrators and resellers, scope usage:read):
curl -s "https://panel.example.com/api/v1/traffic/servers?filter[month]=2026-11&filter[state]=over&sort=-counted" \
-H "Authorization: Bearer YOUR_API_KEY"
GET /api/v1/traffic?month= gives a month's ledger including servers deleted during the month, for billing.
Audit entries#
vm.traffic_notice, vm.throttled, vm.throttle_lifted, vm.throttle_ended, vm.limits_changed, and vm.suspended / vm.unsuspended for traffic.