Metrics
See your traffic. Catch trouble first.
Metrics opens with a one-line verdict, then the detail behind it: a disk forecast that counts down to the day deploys stop, CPU ranked by the app burning it, and privacy-friendly traffic statistics for every site you host. All of it measured on your own server.
| acme-db | is crash-looping: 4 auto-recoveries in 6h | error |
| shop.acme.dev | cron “nightly-backup” last exited 1 | warning |
| Today 09:14 | api.acme.dev deployed | edi |
| Today 03:02 | acme-db restarted automatically | runtime |
| Yesterday 22:40 | shop.acme.dev health recovered | watchdog |
Traffic
The traffic questions site owners actually ask
Per site, DCMS gives you Pulse, top and entry Pages, Referrers by channel, an Audience broken down by country, device, OS and browser, a Slow lane that ranks your slowest paths by p95, and Reliability. Every period is drawn against the one before it, and 365 days are kept.
- Offline geolocationcountries come from a local DB-IP Lite database, so no lookup ever leaves your server.
- Visitors are hashedbefore storage: honest counts without keeping anyone's identity.
- No third-party scriptsnothing to slow your pages down or feed data to ad networks.
- Slow laneranks paths by p95 so you fix what visitors actually feel, not what the average hides.
| Path | p95 |
|---|---|
| /checkout | 1,240 ms |
| /search | 610 ms |
| /product/:slug | 190 ms |
CPU
Find the app burning the box
When the host gets busy, averages hide the culprit. DCMS ranks your instances by the CPU they actually consumed (cumulative CPU·hours or average while running, your choice), so a runaway script has nowhere to hide.
- Cumulative or averageone toggle: total CPU·hours catches the all-day grinder, average catches the short sharp spike.
- Hourly snapshotsof CPU and RAM, kept per instance for 365 days rather than lumped per server.
- Live host metersCPU, RAM and DISK sit in the sidebar on every page of the panel, no report to open.
Host capacity
See the day your disk runs out
Host capacity charts disk used against the line where deploys stop, fits the trend, and tells you in words: growing 340 MB a day, deploys refused in about 26 days. Next to it, a breakdown puts a name on every line of the disk; even the catch-all is labelled, not a mystery gap.
- A countdown, not a pie chartthe forecast is fitted from weeks of samples and stated in days, so you act before deploys stop.
- Cleanup you can trustthe reclaimable figure counts only what a cleanup will actually delete, not Docker's optimistic number, and one click frees it.
- Docker made legibleimages, containers & logs, and build cache & volumes each get their own line instead of hiding in one number.
- Runtime image shareper-instance disk meters even attribute each app's slice of the runtime image it runs on.
Know exactly what your server is doing
Install DCMS and get privacy-friendly analytics plus honest resource numbers for every site you host.