[2.16.1 -> 2.17.0] graphql codegen container gets rejected (ECONNRESET)

[Zerebokep]

Was there any undocumented change regarding connection limits in the 2.17.0 update?

After the update my graphql codegen container instance receives the following error every time I'm generating my graphql code:

FetchError: request to http://hasura:8080/v1/graphql failed, reason: socket hang up

                        reject(new FetchError(`request to ${request.url} failed, reason: ${err.message}`, 'system', err));
          ^
FetchError: request to http://hasura:8080/v1/graphql failed, reason: socket hang up
    at ClientRequest.<anonymous> (/redacted/redacted/node_modules/node-fetch/lib/index.js:1491:11)
    at ClientRequest.emit (node:events:513:28)
    at ClientRequest.emit (node:domain:489:12)
    at Socket.socketOnEnd (node:_http_client:514:9)
    at Socket.emit (node:events:525:35)
    at Socket.emit (node:domain:489:12)
    at endReadableNT (node:internal/streams/readable:1359:12)
    at processTicksAndRejections (node:internal/process/task_queues:82:21) {
  type: 'system',
  errno: 'ECONNRESET',
  code: 'ECONNRESET'
}
error Command failed with exit code 1.

[ajohnson1200]

There were no changes to connection limits in 2.17. Happy to check this out in more detail / try to reproduce if you could provide some additional detail on how / where you're seeing this.

[Zerebokep]

I'll try to share more details asap. So far I've figured out that hasura closes the connection if too much requests happen in a short timeframe, it works fine if I add a 1 second sleep between the requests.

Strangely everything works (even without the 1 second sleep) if I route the requests through my reverse proxy (caddy) instead directly between my two docker containers (hasura & graphql codegen).

#9400 might be related?

[goof03]

Hi!
we have same problem here after an upgrade 16.1 -> 17.1 we cant login anymore…
nodejs api (grapql-request) make :

FetchError: request to http://xx…xx/v1/graphql failed, reason: socket hang up

it's a connection between two pods in a k3s kubernetes cluster.

Downgrading to 16.1 solve the problem

Perhaps a clue, in my team more you computer is powerfull more you can launch request before error …

[motopods]

Downgrading to 16.1 solve the problem, too.
An error occurs if there is no delay between two requests.

[AndRud]

I have the same issue when upgrading from 2.16.1 to 2.17.0 or later.

[YuriyKabanenko]

I have the same problem.
For actions it works fine unlike event triggers

[goof03]

Downgrading node version solve the problem there is a bug in a sub package of graphql lib. I don't remember exacly and i havent check recently if it's fix … sorry

[vitaliytv]

switch from node 20 to node 21 - helped for me