Understanding Sessions

In most cases, setting the Replay sample rates will be all you need to do to begin capturing recording data you care about. For more complex cases, it's helpful to understand how sessions work and how to manually control them.

By default, Replay will immediately start a session when the replayIntegration() integration instance is added to the SDK client. The session will be in either session or buffer mode, depending on your replaysSessionSampleRate and replaysOnErrorSampleRate.

When Replay is initialized, we check the replaysSessionSampleRate. If it's sampled, we'll start recording and sending Replay data immediately.

If replaysOnErrorSampleRate > 0, we'll start recording in buffering mode and keep checking replaysOnErrorSampleRate whenever an error occurs. Once we see that it's sampled, we'll upload the Replay to Sentry and continue recording normally.

If both replaysSessionSampleRate and replaysOnErrorSampleRate are 0, the recording won't start, but you'll be able to record manually.

In session mode, Replay will continuously record and send data to Sentry. After 15 minutes of user inactivity, or a maximum duration of 60 minutes, the session will end and a new session will be initialized based on the rules from the section above.

In buffer mode, Replay will continuously record data, but won't send any of it. It will instead keep up to the last 60 seconds in memory. When an error occurs, replaysOnErrorSampleRate will be checked and if it's sampled, the replay will be uploaded to Sentry and continue recording normally. After 15 minutes of user inactivity, or a maximum duration of 60 minutes, the session will end and the replay will stop.

It is also possible to defer the initialization and loading of the Replay integration. This is helpful to decouple the SDK initialization from Replay configuration. For example, if you have an external sampling service that is not immediately available.

Copied
async function init(sessionSampleRate, errorSampleRate) {
  const client = Sentry.getClient();
  const options = client.getOptions();
  options.replaysSessionSampleRate = sessionSampleRate;
  options.replaysOnErrorSampleRate = errorSampleRate;

  const replay = Sentry.replayIntegration({
    maskAllText: true,
    // additional SDK config, see https://docs.sentry.io/platforms/javascript/session-replay/configuration/
  });

  client.addIntegration(replay);
}

You can manually start a Replay session with:

Copied
Sentry.init({
  dsn: "...",
  // This initializes Replay without starting any session
  replaysSessionSampleRate: 0,
  replaysOnErrorSampleRate: 0,
  integrations: [Sentry.replayIntegration()],
});

// You can access the active replay instance from anywhere in your code like this:
const replay = Sentry.getReplay();

// This starts in `session` mode, regardless of sample rates
replay.start();

// OR

// This starts in `buffer` mode, regardless of sample rates
replay.startBuffering();

This can be used either if both replaysSessionSampleRate and replaysOnErrorSampleRate are 0 and thus no session has been started automatically, or if you previously stopped a session and want to start a new one (see below). start() and startBuffering() will throw an error if a session is currently running.

You can always stop a running session with:

Copied
await replay.stop();

This will flush any pending recording data, stop the replay, and end the session.

You can flush any pending recording data with:

Copied
await replay.flush();

In session mode, this will upload any pending recording data to Sentry. In buffer mode, this will upload any pending recording data to Sentry and then continue recording, the same as when an error is sampled with replaysOnErrorSampleRate.

Note that it's safe to call flush() at any time, even if Session Replay is stopped, in which case, it will do nothing.

There are ways to enable custom sampling if you're interested in tracking a particular action or group, for example, a specific segment of users or a problematic URL.

You can set up replays to be recorded automatically for a particular user group. Here's an example where we want to record session-based replays for all users who are employees. The user object has an isEmployee field to signal employee status.

Copied
const replay = Sentry.replayIntegration();

/**
 * Initialize the Sentry SDK as normal.
 *
 * `replaysSessionSampleRate` is set to its default value of "0.1" so that only ~10% of users are sampled
 * `replaysOnErrorSampleRate` is set to its default value of "1.0" to always record a replay when an error occurs
 */
Sentry.init({
  dsn: "...",
  replaysSessionSampleRate: 0.1,
  replaysOnErrorSampleRate: 1.0,
  integrations: [replay],
});

// ...
// After a user has been authenticated, check if they're an employee
// If they are, we'll call `replay.flush()` to either flush the replay as normal if it's a session-based replay, or if it's buffering due to error sampling, the flush will send the first segment of the replay.

// Since `replaysOnErrorSampleRate` is > 0, we know that a replay has been buffered and `flush()` will flush the contents of that buffer. If only `replaysSessionSampleRate` is > 1, then there is a chance that a replay has not been recording/buffering.
// In this case, you can check that `replay.getReplayId()` returns a value, if it does, it means replay is active and you can call `replay.flush()`, other call `replay.start()` to start recording a new replay.
// ...

// Check if user is an employee, if so, always record a replay
if (loggedInUser.isEmployee) {
  // You can get a reference to the Sentry Replay integration like so:
  const replay = Sentry.getReplay();

  replay.flush();
}

You can also choose to record a replay every time a user lands on a specified URL. This can help developers debug problematic pages. To do this, use the Navigation API, which can listen to navigation events for a specific URL and start the replay recording.

Copied
const replay = Sentry.replayIntegration();

/**
 * Initialize the Sentry SDK as normal.
 *
 * `replaysSessionSampleRate` and `replaysOnErrorSampleRate` are both set to
 * "0" so that we have manual control over session-based replays
 */
Sentry.init({
  dsn: "...",
  replaysSessionSampleRate: 0,
  replaysOnErrorSampleRate: 0,
  integrations: [replay],
});

// Listen to navigation events for when users land on our buggy URL, appropriately named: "/buggy-page/"
// and then start recording a replay.
navigation.addEventListener("navigate", (event) => {
  const url = new URL(event.destination.url);

  if (url.pathname.startsWith("/buggy-page/")) {
    // User navigates to the buggy page, let's start recording
    replay.start();
  }
});

When using replaysOnErrorSampleRate, we "roll the dice" on any Error that is captured & sent to Sentry. If you want to skip capturing a Replay for certain errors, you can use the beforeErrorSampling callback:

Copied
replayIntegration({
  beforeErrorSampling: (error) => {
    return error.message?.includes("drop me");
  },
});

If you return false from this method for a given error, we will not check the error sample rate for this error. If you return true, we will continue to check the error sample rate as normal.

Help improve this content
Our documentation is open source and available on GitHub. Your contributions are welcome, whether fixing a typo (drat!) or suggesting an update ("yeah, this would be better").