My worker wrangler TOML file has following settings:

[[env.production.services]]
binding = "workerA"
service = "worker-a"

[[env.production.services]]
binding = "workerB"
service = "worker-b"

Both services are called like that:

const workerAResponse = await env.workerA.fetch(
  new Request(new URL(`worker-a/${id}`, env.WORKER_A_DOMAIN)),
  {
    headers: request.clone().headers,
  }
);


if (!workerAResponse.ok) { return workerAResponse }

const { public_id } = await workerAResponse.json();

const workerBResponse = await env.workerB.fetch(
  new Request(new URL(`worker-b/${public_id}`, env.WORKER_B_DOMAIN)),
  {
    headers: request.clone().headers,
  }
);

if (!workerBResponse.ok) { return workerBResponse; }

Request to

worker-a

is successful, however a request to` worker-b` is canceled, as

wrangler tail

says:

GET <some-url> - Canceled @ 4/14/2023, 5:03:23 PM

What can possibly go wrong about it? And why worker-b response is canceled whereas worker-a (both are set up literally the same) works just fine. By the way, worker responds with

200

, but due to the

outcome: "canceled"

it is not taken into account ;/

Are you reading the response of Worker B?

in case it fails (not "ok"), I return it. However if it returns 200 I do nothing with the response

You'll want to read the response if it's okay

wow, didnt know I should always read what my worker returns me

It's a quirk specific to Service Bindings and I don't know why - but it'll sit there waiting for you to consume the body and when you don't, that 'response' is cancelled as it wasn't consumed fully.

by the way, should i specify in workerB wrangler.toml that some other worker will do the fetch to this worker?

Not necessary, but if you mean that you don't want anything other than your Worker to be able to fetch it then set

workers_dev = false

and remove any `route`/`routes` keys so that it's only possible to reach it over Service Bindings