What is synthetic monitoring? From uptime checks to browser journeys
Learn what synthetic monitoring measures, how API and browser checks differ from RUM, and how to choose useful assertions and test locations.
At 3 a.m., a checkout endpoint returns HTTP 200, but the response says checkout is unavailable. A status-only check passes. A check that also inspects the business result can catch the failure before the next customer arrives.
Synthetic monitoring runs scheduled requests or scripted journeys from known locations. Each run checks a specific outcome, such as a DNS answer, an API response or a confirmation screen. The result describes what that run could reach and verify from that location.
Decide what a successful run proves
Consider a read-only checkout API. A useful success condition combines the expected HTTP status with an application result, such as available = true. A response of 200 OK with {"available":false} should fail this check. If a reverse proxy returns a branded error page with status 200, a content or schema assertion can catch that too.
Test the failure condition before adding notifications. A run that fails can become a signal for an alert rule. When someone is notified depends on the schedule, timeouts, retries and configured failure conditions. A status code by itself also gives little help in deciding which layer failed.
Choose the smallest check that covers the risk
| Check | What it can establish | What to check next |
|---|---|---|
| ICMP ping or TCP connect | A probe gets an ICMP reply or establishes a TCP connection from its node | Use an application check to verify a database query, login or other operation |
| DNS or TLS | The configured resolver returns the expected answer, or the TLS check verifies the expected identity and certificate validity | Verify the HTTP response and the behavior it represents |
| HTTP or multi-step API | The endpoint returns the expected status and content, or a sequence of requests produces the expected result | Use a browser when the failure depends on rendering or page interaction |
| Scripted browser | The tested page loads, and the scripted actions and UI assertions complete | Use real user monitoring to see how performance varies across actual sessions |
Browser runs also help you compare performance under controlled conditions. Keep the node, browser and script comparable, then look at navigation timing and the resource waterfall where available. A load-only run has no interaction from which to calculate Interaction to Next Paint (INP). An interaction recorded in a script represents that scenario. Field data shows the distribution across actual sessions. Lab and field Web Vitals
Synthetic monitoring and real user monitoring
Synthetic checks can run when nobody is visiting. They are useful when you want to repeat the same checkout path after a release or test a private dependency from a known network.
Real user monitoring (RUM) describes the sessions your instrumentation actually collects. It can expose problems tied to a particular device, geography or interaction that a fixed script never exercises. If RUM shows checkout slowdowns on a particular route, use that finding to choose a browser scenario you can repeat. When a synthetic check first reports a failure, RUM can help you see whether observed users are affected. RUM platform selection guide
Put the test where the failure can happen
A healthy result in one region can coexist with a DNS, routing or CDN problem in another. Run the same check from the locations that matter to the service and keep the per-node results. Several nodes failing at once can support escalation for a widespread outage. A rule that requires several failures can miss a genuine outage isolated to one region, so give regional symptoms their own policy where the impact warrants it.
Guance offers managed nodes and self-built synthetic test nodes. For a private API, install DataKit on a host with access to the target and the required control and ingestion endpoints, then configure the dialtesting collector. Private addresses also require enabling internal-network tasks in that collector. Check node and task support in your workspace. A probe inside your network measures a different path from an external customer.
Set up one check you can investigate
Start with a read-only request that represents a real business decision. In a Guance HTTP synthetic test, configure the request and availability conditions. Use the default conditions when they express the rule. Use Pipeline script mode when you need to parse the response and evaluate a business field. Try one expected success and one controlled failure, then inspect a scheduled result from the node you selected. The initial connection test runs independently of that node selection.
If the failure can only happen after JavaScript runs, move the assertion into a browser synthetic test. For checkout, an example is adding a test item and checking that the cart shows the expected item and total. Use a test account and a sandbox for writes. When a run fails, use its time, node and error details to narrow the scope, then compare the corresponding logs or traces where they are available.
Choose the interval from the time you can tolerate before detection, the alert rule and the request cost. Where a configured marker identifies probe traffic, filter it from user or business metrics that would otherwise count it. Count the executions from each node and any retries when assessing load. Uptime monitoring (Simplified Chinese)
Contact us
Join the community
to join the community
Try Guance
Start online and pay only for what you use.
Get startedChoose a Guance plan
