Slowdown after deploying Sentry Nuxt SSR

[cernymatej]

Is there an existing issue for this?

• I have checked for existing issues https://github.com/getsentry/sentry-javascript/issues
• I have reviewed the documentation https://docs.sentry.io/
• I am using the latest SDK release https://github.com/getsentry/sentry-javascript/releases

How do you use Sentry?

Self-hosted/on-premise

Which SDK are you using?

@sentry/nuxt

SDK Version

9.30.0

Framework Version

Nuxt 3.17.5

Link to Sentry event

No response

Reproduction Example/SDK Setup

I've recently deployed the Sentry integration to my nuxt projects and have noticed a significant performance problem, where the response time worsened by around 100ms.

Is this a known issue? I don't have a reproduction yet as I haven't looked into it in too much detail just yet. All I know is that when I don't use the sentry module and don't start the server using node --import ./.output/server/sentry.server.config.mjs .output/server/index.mjs, the response time for each request is much better.

Expected Result

No noticeable performance decrease

Actual Result

A +- 100ms additional latency in response time

[linear]

JS-738 Slowdown after deploying Sentry Nuxt SSR

[s1gr1d]

Hello, can you share a small reproduction example? Or most importantly, share your Sentry.init options from client and server?

[cernymatej]

Sure thing! Here are the config files:

// sentry.client.config.ts

import { useRuntimeConfig } from '#imports'
import { init, replayIntegration, consoleLoggingIntegration } from '@sentry/nuxt'

const runtimeConfig = useRuntimeConfig()

init({
    dsn: runtimeConfig.public.sentry.dsn,
    environment: runtimeConfig.public.sentry.environment || 'unknown',
    sendDefaultPii: true,
    integrations: [
        replayIntegration(),
        consoleLoggingIntegration({ levels: ['error', 'warn'] }),
    ],
    tracesSampleRate: 1.0,
    tracePropagationTargets: ['*'],
    replaysSessionSampleRate: 0,
    replaysOnErrorSampleRate: 1.0,
    _experiments: {
        enableLogs: true,
    },
})

// sentry.server.config.ts

import { init, consoleLoggingIntegration, captureMessage } from '@sentry/nuxt'

// a temporary workaround until our self-hosted Sentry supports the new experimental logs
const originalConsoleError = console.error
console.error = (...args: any[]) => {
    captureMessage(args.join(' '), 'error')
    originalConsoleError.apply(console, args)
}

if (process.env.NUXT_PUBLIC_SENTRY_DSN) {
    init({
        dsn: process.env.NUXT_PUBLIC_SENTRY_DSN,
        environment: process.env.NUXT_PUBLIC_SENTRY_ENVIRONMENT || 'unknown',
        integrations: [
            consoleLoggingIntegration({ levels: ['error', 'warn'] }),
        ],
        tracesSampleRate: 1.0,
        _experiments: {
            enableLogs: true,
        },
    })
}

// nuxt.config.ts

modules: [
    '@sentry/nuxt/module',
],

I haven't been able to create a minimal reproducible example yet. We first noticed the issue in a production project, and I can't see any difference in response time when I strip it down. I can, however, reliably reproduce the slowdown in the full project even locally. Could you give me some hints on where to start looking? My usual divide-and-conquer strategy for creating a reproduction didn't work that well this time 😅

[s1gr1d]

Where do you deploy your application? What could increase the response time is the awaiting of the flushing (code here). But this is only happening after a response is already sent. It would just increase the runtime of a serverless function by a maximum of 2 seconds to be able to still send the event to Sentry.

Can you also try minimizing your tracesSampleRate to a very low value like 0.1 or even 0 to see if there is a significant change. Generally, this value is recommended to be lower than 1 in production environments, depending on your use case.