Dated host snapshots
See recent CPU, load, memory, root disk, selected service states, and a bounded process summary. A missed report appears as a connection gap, not proof of an outage.
Optional server evidence for MetricPoints
See what the server was doing when a client site failed.
Public checks show what visitors can reach. The optional Linux agent adds dated host, service, and process evidence to help your agency investigate one failing site or several on the same server.
The agent is optional on every plan, including free Developer. Connecting a server requires a MetricPoints account and administrator access to the Linux host.
Illustrative investigation
Two client sites fail on one host
-
1
Public checks detect the failure
Recent outside results show which sites are affected and when the problem began.
-
2
The agent supplies host context
A recent report may show a stopped web service, resource pressure, or a PHP-FPM queue at the same time.
-
3
Your team gets one explanation
When the site links and timing are clear, MetricPoints can show shared-server evidence with the affected sites and their incident history.
Server evidence helps investigation; a matching snapshot does not prove the host caused a public failure.
What the agent adds
Server context beside the site incident.
Use the same All Sites and Site Health views to investigate what visitors experienced and what the connected server reported.
Incident-triggered diagnosis
For eligible current site incidents, collect a read-only snapshot of web and PHP-FPM context. Available queue evidence is shown with supporting and contradictory signals.
Shared-server visibility
Match reported virtual-host domains to agency sites provisionally. When multiple current site failures have fresh, unambiguous links to one server, see their shared context together.
Connection and compatibility
Connect from your Servers page.
- Choose a server record and generate a short-lived install command in MetricPoints.
- Run it with administrator access on the Linux host. The agent reports over outbound HTTPS; no inbound port is needed.
- Check the first report, then review provisional matches between that host and your client sites.
Reporting and diagnosis have been compatibility-tested on Ubuntu 20.04, 22.04, 24.04, and 26.04, and AlmaLinux 8 and 9, on x86_64 or aarch64 with systemd and required host commands. Check extended security maintenance before using Ubuntu 20.04 in production.
Open Servers to connect →Safe by default
Evidence first. Your team stays in control.
- Routine reporting is read-only. It sends bounded server and service summaries, not raw access-log lines, request bodies, or credentials.
- Public checks still determine site failure. High CPU or a stale agent report alone does not create a customer incident or trigger a repair.
- Server actions are a separate controlled pilot. A PHP-FPM restart requires an eligible non-client-facing site incident, explicit approval, local permission, and deployment safeguards; automatic restarts are unavailable.
External monitoring works without an agent or WordPress. Connect only the servers where additional host evidence would help your agency investigate.