In v8, manual initialization of the SDK will work as follows.
import * as Sentry from '@sentry/browser'; Sentry.init({ dsn: 'xxx', });
This will initialize the SDK with all defaults & make it the currently active Sentry instance. This will continue to use the existing global, isolation & current scope, and just bind the client to it.
In an environment with multiple execution contexts (e.g. Node), you can setup multiple clients that are active for different contexts, like this:
import * as Sentry from '@sentry/node'; // Sets up the _default_ client Sentry.init({ dsn: 'xxx', }); // One execution context with client A Sentry.withScope(() => { const clientA = new Client(); Sentry.setCurrentClient(clientA); // binds this client to the current execution context only! clientA.init(); }); // One execution context with client B Sentry.withScope(() => { const clientB = new Client(); Sentry.setCurrentClient(clientB); // binds this client to the current execution context only! clientB.init(); });
In environments without execution contexts, like the browser, you can only ever have a single active client. You can, however, create further clients and use them manually:
// Default client - this is used everywhere Sentry.init({ dsn: 'xxx', }); // Setup a manual client const clientA = new Client(); const scope = new Scope(); scope.setClient(clientA); // You can capture exceptions manually for this client like this: scope.captureException();
This is also necessary e.g. if you have a browser extension or some other code that runs in a shared environment.