Diagnostics and location pools
When a monitor fails from one place but not another, you want to see the network from that place. Diagnostics runs a one-off ping, traceroute, MTR, DNS lookup or HTTP timing against one of your hosts from one or several monitoring locations at the same time and shows the answers side by side. This page also covers location pools on the monitor form and the Regions and networks card on a monitor's page.
Opening the page
Diagnostics is in the panel menu, right below Servers (/app/diagnostics). Everyone who can view monitors sees the page and the history; starting a run needs the Add and edit monitors permission.
On an upgraded installation the page says Diagnostics are not set up yet until the administrator has opened db_update.php once.
Starting a run
The New diagnostic card has:
| Field | Meaning |
|---|---|
| Tool | Ping, Traceroute, MTR, DNS lookup or HTTP timing |
| Target | A host name, IP address or URL (up to 500 characters). The list suggests the hosts of your monitors and servers |
| Packets | Ping and MTR: 1–20, default 5 |
| Max hops | Traceroute and MTR: 3–30, default 20 |
| Record type | DNS lookup: A, AAAA, CNAME, MX, TXT, NS, SOA, PTR, SRV or CAA |
| Resolver | DNS lookup: a resolver to ask, such as 1.1.1.1 or 8.8.8.8. Empty = each location's own resolver |
| Method | HTTP timing: GET or HEAD |
| Locations | Main server plus every probe that is online right now. The card shows how many you may pick |
Press Run. The result page fills in as each location answers (it refreshes itself every few seconds), and Run again starts the same diagnostic once more.
Which targets are allowed
Diagnostics are for your own infrastructure, not a free scanning service:
- The target must be the host name or IP address of one of your organisation's monitors or servers (the host of a monitor's URL, or a server's hostname or IP). Anything else is refused with You can only diagnose the hosts of your own monitors and servers. Platform administrators may diagnose any public host.
- Every target, and a custom DNS resolver, goes through the same SSRF guard as monitors: private, reserved and cloud-metadata addresses are refused.
- A DNS lookup of a name is allowed even if the name does not resolve; an IP target is still checked by the guard.
- Commands only ever receive the resolved IP address, never the text you typed.
Limits and plans
- Rate limit: an organisation may start 20 runs per hour by default (the card shows the number). The administrator can change it with the setting
diag_rate_per_hour. - Plans with one monitoring location run diagnostics from the main server only (Your plan runs diagnostics from the main server only. Upgrade to compare locations.).
- Other plans may pick as many locations as their plan's location limit allows, at most 10 per run (10 for an unlimited plan).
- Only probes that are online, public and not being drained are offered.
The tools
| Tool | What runs on the location | What you see |
|---|---|---|
| Ping | ping (Linux: 0.5 s apart, 2 s timeout per packet; Windows: 2 s timeout) | Replies, packet loss, min / avg / max, jitter and TTL |
| Traceroute | traceroute with 3 queries per hop (or tracepath when traceroute is missing); tracert on Windows | One row per hop: host, IP, loss, average and each round-trip time; Destination reached or not reached within the hop limit |
| MTR | mtr --report --json (Linux only, mtr must be installed) | One row per hop: loss, average and best / worst |
| DNS lookup | The built-in DNS client (the same one DNS monitors use) | Record type, resolver and the sorted answer |
| HTTP timing | The same request an HTTP monitor makes, following redirects | Status code, DNS, Connect, TLS, First byte and Total times, size, Server header, redirect chain and certificate expiry |
Every location card also shows the IP used, the program and platform, how long it took and, under Raw output, the program's own output.
A ping counts as failed when no reply came back; an HTTP timing is successful whenever the server answered with any status code. When a program is not installed or PHP may not run programs on that location (common on shared hosting) the card says not available instead of failed, for example MTR is not available on this location. Run diagnostics from a probe on a VPS for full network tools.
Comparing DNS answers
When a DNS lookup ran from two or more locations, the top of the result page says either Every location got the same answer or The answers differ between locations with each answer and the locations that received it. Use this to see a propagation delay or a geo-DNS answer at a glance.
When a location does not answer
A location that has not answered after 6 minutes (offline probe, probe without diagnostics support, no worker running) ends as No answer from this location in time. A run is finished when every location has an answer or has timed out.
History
The History card lists the last 50 runs of your organisation: tool, target, locations, status (running, done, or N failed), who started it and when. Runs are kept for 30 days.
Location pools (monitor form)
Besides fixed locations, the monitor form has a Location pools card (shown when the administrator runs probes with a region):
- Any probe in a region: tick a region and each check runs on the least busy online probe of that region. The stored location is
region:<Region>, for exampleregion:Europe. - Automatic spread: N probes from N different regions (2 up to the number of regions). Each check picks the least busy probe of every region not used yet and keeps the N least busy of those. Stored as
auto:N.
Rules:
- A region pool counts as one location of your plan;
auto:Ncounts as N. - Only public probes (and the main server's own region) take part in pools.
- "Least busy" is the number of jobs waiting for a probe plus the queue it reported in its last heartbeat, divided by its max jobs per pull times its weight.
- The result is stored with the code of the probe that really ran the check, so the per-location history shows the actual probes.
- If no probe of a region is available, the check runs on the main server.
The API accepts the same tokens in the locations field:
"locations": "core,region:Europe,auto:3"
Failover
A fixed probe that is offline or being drained is replaced, for each check, by the least busy online probe of the same region, or by the main server when the region has none. Checks never stop because a probe went away. See Probes.
Regions and networks card
On a monitor's page, once its results come from at least two locations, the Regions and networks card compares the last 24 h or 7 days:
- By location, By network (ISP / ASN) and By region: failed / total checks, the failure percentage, and the average and p95 response time of the successful checks.
- An ISP-specific problem note appears when, in the newest result of each location, only probes of one ASN fail while probes on other networks pass: Only the probes on … fail … The target may be unreachable from that provider only.
- A Regional problem note appears when only the locations of one region fail while the other regions pass.
The ISP and ASN of each probe are filled in by the administrator (Admin → Probes, with an RDAP lookup). The card links to Run a traceroute from several locations, which opens Diagnostics with the monitor's host filled in.
One click helps us improve these docs.

