Closed
Description
Important Details
How are you running Sentry?
Vercel
Description
import { logEvent } from '@/modules/core/amplitude/amplitudeServerClient';
import {
AMPLITUDE_API_ENDPOINTS,
AMPLITUDE_EVENTS,
} from '@/modules/core/amplitude/events';
import { filterExternalAbsoluteUrl } from '@/modules/core/js/url';
import { createLogger } from '@/modules/core/logging/logger';
import { FLUSH_TIMEOUT } from '@/modules/core/sentry/config';
import { configureReq } from '@/modules/core/sentry/server';
import * as Sentry from '@sentry/nextjs';
import appendQueryParameter from 'append-query';
import {
NextApiRequest,
NextApiResponse,
} from 'next';
const fileLabel = 'api/preview';
const logger = createLogger({
fileLabel,
});
/**
* Key to use in order to force disable "auto preview mode".
*
* If a page is loaded with noAutoPreviewMode=true in query parameter, then it won't try to enable the Preview mode, even if it's disabled.
*
* @example ?noAutoPreviewMode=true
*/
export const NO_AUTO_PREVIEW_MODE_KEY = 'noAutoPreviewMode';
type EndpointRequestQuery = {
/**
* Whether to start/stop the Preview Mode.
*
* @example ?stop=true Will stop the preview mode.
* @example ?stop=false Will start the preview mode.
* @default ?stop=false
*/
stop: string;
/**
* Url to redirect to once the preview mode has been started/stopped.
*
* @example ?redirectTo=/en
* @example ?redirectTo=/fr/solutions
* @default ?redirectTo=/
*/
redirectTo: string;
}
type EndpointRequest = NextApiRequest & {
query: EndpointRequestQuery;
};
/**
* Preview Mode API.
*
* Enables and disables preview mode.
*
* The official example uses a security token to enable the preview mode, we don't.
* This is a choice, as we don't need/want to protect our preview mode.
* Protecting the preview mode makes most sense when this mode can be used in production, so that you can preview content served by Next.js from a CMS/tool of your choice.
* Thus, it's strongly related to how you're planning on using it, and we decided to keep it simpler, by not using any kind of security.
*
* @param req
* @param res
* @method GET
*
* @see https://nextjs.org/docs/advanced-features/preview-mode#step-1-create-and-access-a-preview-api-route
* @see https://nextjs.org/docs/advanced-features/preview-mode#clear-the-preview-mode-cookies
*/
export const preview = async (req: EndpointRequest, res: NextApiResponse): Promise<void> => {
try {
configureReq(req, { fileLabel });
await logEvent(AMPLITUDE_EVENTS.API_INVOKED, null, {
apiEndpoint: AMPLITUDE_API_ENDPOINTS.PREVIEW,
});
const {
stop = 'false',
redirectTo = '/',
}: EndpointRequestQuery = req.query;
// Add NO_AUTO_PREVIEW_MODE_KEY parameter to query, to avoid running into infinite loops if the Preview mode couldn't start
// Useful when the cookie created by Next.js cannot be written (Incognito mode)
const safeRedirectUrl = appendQueryParameter(filterExternalAbsoluteUrl(redirectTo as string), `${NO_AUTO_PREVIEW_MODE_KEY}=true`);
// XXX We don't want to enable preview mode for the production stage, it's only allowed for non-production stages
// It's allowed during development for testing purpose
// It's allowed during staging because this stage is being used as a "preview environment"
if (process.env.NEXT_PUBLIC_APP_STAGE !== 'production') {
if (stop === 'true') {
res.clearPreviewData();
logger.info('Preview mode stopped');
} else {
res.setPreviewData({});
logger.info('Preview mode enabled');
}
} else {
logger.error('Preview mode is not allowed in production');
Sentry.captureMessage('Preview mode is not allowed in production', Sentry.Severity.Warning);
}
// It's necessary to flush all events because Vercel runs on AWS Lambda, see https://vercel.com/docs/platform/limits#streaming-responses
await Sentry.flush(FLUSH_TIMEOUT);
res.writeHead(307, { Location: safeRedirectUrl });
res.end();
} catch (e) {
Sentry.captureException(e);
logger.error(e);
// It's necessary to flush all events because Vercel runs on AWS Lambda, see https://vercel.com/docs/platform/limits#streaming-responses
await Sentry.flush(FLUSH_TIMEOUT);
res.json({
error: true,
message: process.env.NEXT_PUBLIC_APP_STAGE === 'production' ? undefined : e.message,
});
}
};
export default preview;When await Sentry.flush(2000); is executed, if it times out, Sentry throws an exception. The exception value is false.
From the TS doc (sdk.d.ts):
/**
* A promise that resolves when all current events have been sent.
* If you provide a timeout and the queue takes longer to drain the promise returns false.
*
* @param timeout Maximum time in ms the client should wait.
*/
export declare function flush(timeout?: number): Promise<boolean>;It should return false, not throw false.
This is broken, it shouldn't throw an exception when timing out, as it breaks the code.
It's sort of a shame that the tool meant to help devs make their app more robust actually breaks them.
When encapsulating with try/catch it works fine.
Using "@sentry/nextjs": "6.6.0". Updating to 6.7.2 to see if it solves the issue.
Steps to Reproduce
- Call
await Sentry.flush(1);in a Next.js API endpoint and deploy to Vercel.
See PR at UnlyEd/next-right-now#370 (migrating from @sentry/node to @sentry/next.js)
What you expected to happen
It shouldn't throw an exception.