[PPR] TypeError: Response.clone: Body has already been consumed.

[steve-marmalade]

@wyattjoh , @ztanner

Testing out PPR and PFPR and it seems to be working well for the most part (nice work). However, in one of my routes I am seeing a 500 for the prefetch RSC GET request. Clicking a Link to those pages causes a full page load, but it does ultimately load the content.

⨯ TypeError: Response.clone: Body has already been consumed.
    at webidl.errors.exception (node:internal/deps/undici/undici:3384:14)
    at _Response.clone (node:internal/deps/undici/undici:8876:31)
    at /var/task/.next/server/chunks/8875.js:2:11741
    at process.processTicksAndRejections (node:internal/process/task_queues:95:5)
    at async /var/task/.next/server/chunks/8047.js:1:182603
    at async K (/var/task/.next/server/chunks/8047.js:1:184526)
    at async /var/task/.next/server/chunks/4498.js:166:33
    at async v (/var/task/.next/server/app/path/to/my/profile/page.js:1:4306) {
  digest: '1674423162',
  page: '/path/to/my/profile'
}

I'm not opening an Issue yet as I don't have a clear way to reproduce (I don't know what's different about this page from others that are working). Happy to help investigate this if you can give me some pointers on what to try / where to look. Also happy to share a Vercel preview build URL / create a Replay, if that's helpful.

[steve-marmalade]

Fixed in #70082 🎉

[OZZlE]

@steve-marmalade can we expect the fix to be made on v14 stable version any time soon? Our entire website is currently broken, we have lots of logging in the app but this is the only reported error I can find

[mogzol]

@OZZlE should be fixed properly with this PR which is merged, so I would expect the next v14 release to include it

[rakhi2104]

hey @steve-marmalade,

We did get this working when we downgraded our NodeJS Docker image to 18-alpine and sticking to Next @ 14.2.15

[beckerei]

I could not observer the behaviour when downgrading to node 20.16 which does not throw the error together with next 14.2.14.
Node 20.17 contains a memory leak in fetch, which is fixed in 20.18. But 20.18 and also 22.9 will are effected by the problem. (we run node alpine as well)
I'm unsure how node version and the problem in next are connected. We are not using PPR at all.

[OZZlE]

why is this marked as fixed when the issue obviously still exists in latest stable version e.g. Next 14, sure it works in node.js 18 but then that should be required or patched in next

[sannajammeh]

Appears this error is now in all our Vercel deployments. Happened overnight with no changes to git.

[Eliotboutet]

Same for us here. Do we have any visibility on the issue resolution ?

[sannajammeh]

Can confirm this issue is due to NodeJS 20.18 which was gradually rolled out as one by one our deployments stopped building. More specifically it's a bug inside Undici 6.19. 6.20 is fixes the issue but not included in Node 20.18. We were able to fix build behaviour by polyfilling fetch to Undici 6.20.

Short term fix is to Downgrade affected projects to Node JS 18.x by specifying this in your engines field in package.json. This fixes the issue. Then we wait for a new NodeJS patch and Vercel update