Skip to content

Real user monitoring

Page loads, Web Vitals, requests and JavaScript errors from your users' browsers, with issues, source maps and session timelines, stored as ordinary traces, logs and metrics.

Updated View as Markdown

Real user monitoring (RUM) measures your web applications as your users experience them. Add the EZGH browser SDK to a web app, and every page load, route change, interaction, network request, Core Web Vital and JavaScript error from real browsers arrives in your project’s Observability, as ordinary traces, logs and metrics in the app’s environment.

  • Performance: LCP, INP, CLS, FCP and TTFB by route, browser, device and country.
  • Errors as issues: symbolicated with your source maps, grouped into issues you resolve or ignore, and reopened when a newer release brings them back.
  • One trace from click to database: when you allow it, requests to your APIs carry traceparent, so the browser span and your backend’s spans are one trace.
  • Session timelines: from an error or a slow view, see that session’s views, interactions, requests and errors in order.
  • Small and safe: the core SDK is under 25 KB gzipped, never breaks your page, stores no IP addresses and captures no page text.
  • Billed as Observability data: browser data is billed as logs and traces ingested bytes, with no RUM-specific price. The Web Vitals, view, error and long-task metrics are generated by EZGH from your view events and errors, and are free, as span metrics are. Source maps, symbolication and issues cost nothing. Errors dropped by spike protection aren’t billed.
The Browser apps page for the production environment: three applications (admin-portal, marketing-site and storefront) with their environments, token count, and p75 LCP, INP and CLS over the last 24 hours with good and needs-improvement dots, and errors per 100 views, with an Add browser app button.The Browser apps page for the production environment: three applications (admin-portal, marketing-site and storefront) with their environments, token count, and p75 LCP, INP and CLS over the last 24 hours with good and needs-improvement dots, and errors per 100 views, with an Add browser app button.

Concepts

Term Meaning
Browser app A web application sending RUM data, named by its application, such as storefront. The console groups client tokens by it; it isn’t a resource of its own.
Client token A publishable credential (ezgh_ct_…) for one environment and application, with allowed origins and limits. It has an ID, a ctk_ slug and a resource name. See Client tokens.
Session One user’s visit: a random ID shared across tabs of the same origin. It ends after 30 minutes of inactivity or 4 hours.
View One page or route: a full page load or a single-page app’s route change, named by its route template, such as /users/:id.
Release Your build’s version: release in init, stored as service.version.
Debug ID A UUID linking a built JavaScript file to its source map. See Source maps.
Issue One kind of error in one environment and application, with a state: unresolved, resolved or ignored. It has an ID, an iss_ slug and a resource name. See Issues and alerts.

How it works

  1. The SDK sends OpenTelemetry (OTLP JSON) to https://obs-browser.ezghcloud.com, or to your own tunnel.
  2. EZGH checks the token, the page’s origin and the token’s limits; adds the country, region and city from the user’s IP address, then discards the address; and applies your dataset’s transform rules.
  3. Errors are symbolicated with your source maps and grouped into issues, usually within 30 seconds.
  4. EZGH generates the RUM metrics from view events and errors (browsers send no metrics). Everything is searchable as logs and traces, the metrics are in PromQL, and the RUM pages in the console read the same data.

Browser data goes to the dataset of the token’s environment and is kept by that dataset’s retention class, like any other logs and traces.

In the console

Open Observability. The Real user monitoring group has three pages:

  • Browser apps: one row per application, with its environments, tokens and 24-hour p75 Web Vitals and error rate. Add browser app creates a client token, and each app has Overview, Tokens, Source maps and Setup tabs. Setup shows the install code for npm, React, Next.js, Vue, Angular and a script tag with the token filled in, the build plugin’s configuration with your organization’s and project’s IDs, and tunnel handlers.

  • Performance: p75 Web Vitals with good, needs-improvement and poor bands, their trends, a table of views by route, and breakdowns by browser, device and country.

    The Performance page for storefront in production over 24 hours: p75 cards for LCP and INP (needs improvement) and CLS, FCP and TTFB (good), each with the share of good views and its thresholds; an LCP p75 chart between the good and poor lines; the views table by route with views, LCP, INP, CLS, errors per 100 views and long tasks; and the breakdown by browser.The Performance page for storefront in production over 24 hours: p75 cards for LCP and INP (needs improvement) and CLS, FCP and TTFB (good), each with the share of good views and its thresholds; an LCP p75 chart between the good and poor lines; the views table by route with views, LCP, INP, CLS, errors per 100 views and long tasks; and the breakdown by browser.
  • Issues: the issues of an environment, by state, with each issue’s stack trace, breadcrumbs, breakdowns, events and session timelines.

Without the actions to change client tokens, source maps or issues, their pages are read-only.

Access

Action Allows Access level Policies
obs.clientTokens.list Listing client tokens List ObsFullAccess, ObsReadOnlyAccess, ReadOnlyAccess
obs.clientTokens.get Viewing a client token, its value included Read ObsFullAccess, ObsReadOnlyAccess, ReadOnlyAccess
obs.clientTokens.create Creating client tokens Write ObsFullAccess
obs.clientTokens.update Changing a token’s name, origins, tunnel addresses and limits Write ObsFullAccess
obs.clientTokens.delete Deleting client tokens Write ObsFullAccess
obs.sourceMaps.upload Uploading source maps Write ObsFullAccess, ObsSourceMapUploadAccess
obs.sourceMaps.list Listing uploaded releases and their files List ObsFullAccess, ObsReadOnlyAccess, ReadOnlyAccess, ObsSourceMapUploadAccess
obs.sourceMaps.delete Deleting a release’s source maps Write ObsFullAccess
obs.issues.list Listing a dataset’s issues List ObsFullAccess, ObsReadOnlyAccess, ReadOnlyAccess
obs.issues.get Viewing an issue, its breakdowns and its events Read ObsFullAccess, ObsReadOnlyAccess, ReadOnlyAccess
obs.issues.update Resolving, ignoring and reopening issues Write ObsFullAccess
  • Creating a token with an environment that has no dataset yet also needs obs.datasets.create.
  • Every issue read and change also needs obs.issues.list on the issue’s dataset, and an issue’s events need obs.logs.search there too: they’re log records.
  • A session timeline needs obs.logs.search and obs.traces.search on the dataset.
  • ObsSourceMapUploadAccess holds only obs.sourceMaps.upload and obs.sourceMaps.list (with organizations.get, projects.list and projects.get): attach it to the bot your CI uses. See Source maps.

Trails

Creating, changing and deleting client tokens, uploading and deleting source maps, and changing an issue’s state are recorded in Trails with eventSource obs.ezghcloud.com, and so are the reads: CreateClientToken, UpdateClientToken, DeleteClientToken, GetClientToken, ListClientTokens, UploadSourceMaps, ListSourceMaps, DeleteSourceMaps, ListIssues, GetIssue and UpdateIssue.

  • A token’s value never appears in Trails. Token events record its name, dataset, application and how many origins it has; source-map uploads, the application, release and file count; issue changes, the new state.
  • Source-map chunk uploads are recorded only when they’re refused (CheckSourceMapChunks, UploadSourceMapChunk): the commit, UploadSourceMaps, records the upload.
  • Data sent from browsers, issue breakdowns, issue events and session timelines aren’t recorded.
Navigation

Type to search…

↑↓ navigate↵ selectEsc close