usage reporting failing with following stacktrace (core v0.8.3)

[tomasstrejcek]

POST https://app.graphql-hive.com/usage failed (7ms). The "listener" argument must be of type function. Received an instance of Object"}
TypeError [ERR_INVALID_ARG_TYPE]: The "listener" argument must be of type function. Received an instance of Object"}
at checkListener (node:events:275:3)"}
at ClientRequest.once (node:events:652:3)"}
at new ClientRequest (node:_http_client:248:10)"}
at request (node:https:381:10)"}
at /usr/app/node_modules/@opentelemetry/instrumentation-http/build/src/http.js:513:41"}
at safeExecuteInTheMiddle (/usr/app/node_modules/@opentelemetry/instrumentation/build/src/utils.js:28:18)"}
at /usr/app/node_modules/@opentelemetry/instrumentation-http/build/src/http.js:505:78"}
at AsyncLocalStorage.run (node:internal/async_local_storage/async_hooks:91:14)"}
at SentryContextManager.with (/usr/app/node_modules/@opentelemetry/context-async-hooks/build/src/AsyncLocalStorageContextManager.js:33:40)"}
at SentryContextManager.with (file:///usr/app/node_modules/@sentry/node/node_modules/@sentry/opentelemetry/build/esm/index.js:1475:24)"}

our implementation:

const hive = createHive({
    enabled: isStaging || isProd,
    debug: isTest || isDevelopment,
    token: config.get<string>('hive.token'),
    usage: {
      exclude: ['sentry_healthcheck'],
      clientInfo(context: MercuriusContext): any {
        const name = context.reply.request.headers['apollographql-client-name']
        const version = context.reply.request.headers['apollographql-client-version']

        if (name && version && typeof name === 'string' && typeof version === 'string') {
          return { name, version }
        }

        return null
      }
    },
    agent: {
      logger: {
        info: (str: string) => {
          logger.info({ str }, 'hive info log')
        },
        error: (str: string, obj: any) => {
          logger.fatal({ obj, str }, 'hive error log')
        }
      }
    }
  })

secondary issue is that the error is passed to the logger so each line is logged separatelny but that i can live with :)

overall usage is still present in the app, but i see few of these errors in the past few days

[kamilkisiela]

Hey, nothing in the stack trace points to Hive. Why do you think it’s related to usage reporting?

[tomasstrejcek]

well it is logged by the hive agent logger, maybe it more points to the opentelemetry dependency? should I raise it there and get back to you?

thanks

[kamilkisiela]

It was logged by the hive logger, most likely because of a try-cache block, but the error is related to your code I think, specifically to the otel instrumentator for the node's http lib.

[tomasstrejcek]

alright, I will look into it again

[tomasstrejcek]

we only managed to resolve this by massdowngrade our package lock and return few days into past, but the issue is not probably related to hive, current suspect in somewhere in sentry packages

thank you

[tomasstrejcek]

@kamilkisiela well, bad news/good news scenario, figured out the culprit
in v 0.8.3 there was a change in how fetch (or what fetch implementation) is used, which breaks when other instrumentations of the http layer (opentelemetry, sentry etc are introduced)

fix is kinda straightforward, when creating hive instance, pass { agent: { fetch: fetch } } and everything is ok

(i am passing ofcourse standard node fetch, not some of the many os implementations)

[tomasstrejcek]

so maybe docs update is in order? should I send PR? possibly somewhere here: https://the-guild.dev/graphql/hive/docs/api-reference/client

[mvantellingen]

Hitting the same issue with createSupergraphManager from @graphql-hive/apollo which depends on the http-client module from @graphql-hive/core. Unfortunately we can't pass an agent for that use-case

[mvantellingen]

Solved this issue for our projects by rewriting the supergraph manager to use native node-fetch, see https://gist.github.com/mvantellingen/9bfeb69df502911b5b508e5982916b18

[dotansimha]

Thanks @mvantellingen @tomasstrejcek for the additional info. We are looking into that!

[n1ru4l]

@mvantellingen Can you share which prometheus and/or Sentry configuration you are using causes these issues as well as the Node.js version?