"CODE_MOVED for unknown code block" bug?

Weird one - just got this apropos of nothing - wasn't even sending requests to the worker, but it was running. Assume a bug?

MityaOP•1/22/24, 3:10 PM

This is while running locally with --remote--remote, incidentally.

MrBBot•1/22/24, 4:32 PM

Hey!

This message is safe to ignore, see https://github.com/cloudflare/workerd/issues/1472#issuecomment-1850322438. We'll make sure this gets hidden from logs.

GitHub

Docker error: CODE_MOVED for unknown code block? · Issue #1472 · cl...

Running wrangler dev --port=8000 inside Docker with: /* eslint-disable */ import { Router } from 'itty-router' const router = Router() const routes = JSON.parse(ROUTES) const proxyTo = host...

MityaOP•1/22/24, 4:49 PM

Thanks for the reply. However in my case it led to all requests spitting out 400s. I had to restart Wrangler/the Worker and all was well again (requests then spat out 200s or whatever they were supposed to.)

MrBBot•1/23/24, 12:10 PM

Hmmm ok, do you have code in your /misc/checkSession/misc/checkSession that could return a 400 response? Is it possible bad requests are being sent for another reason?

MityaOP•1/24/24, 1:49 PM

No - that endpoint returns either 200 or 401. I don't think it's any other part of my code as the problem went away when I restarted the worker. I'll post back here when it next happens and see what I can identify.

MityaOP•1/26/24, 1:38 PM

@MrBBot Did some more digging thing. Definitely something up with Wrangler/CF. It seems to happen when my back-end is idle for a while i.e. receiving no requests. Apropos of nothing it then gives this error, and any subsequent requests give a 400. If I restart Wrangler, the problem goes away and all requests work again.

MityaOP•1/26/24, 3:08 PM

Should I be worried about this happening on live workers, or is it only on local dev?

MityaOP•1/26/24, 3:08 PM

I'm working with --remote--remote active, so it is running on your network, even when developing.

MMitya @MrBBot Did some more digging thing. Definitely something up with Wrangler/CF. I...

MrBBot•1/29/24, 11:48 AM

Thanks for the additional information. I wouldn't worry about this for live workers. Even with --remote--remote, we still run a local instance of the workers runtime for proxying requests to the remote server, so I'm guessing these errors are coming from that. I'll see if I can find some time to reproduce this, but if you could put together a reliable reprodution that would be super helpful.

MityaOP•1/29/24, 1:53 PM

Thanks for the reply. As far as I can tell the repro steps are:

basic CF Worker example
I've attached my
```
package.json
```
```
package.json
```
and dev.tomldev.toml
Leave Worker idle for some time
Console shows output I sent in previous message
Browser dev tools confirms it's getting unprocessable responses (see screenshot)
When right-clicking on one of those requests and choosing "Open in new tab", I get the attached error output

dev.toml296B

package.json792B

MityaOP•1/29/24, 1:53 PM

As previously mentioned, this persists until I quit the worker and restart it.

MrBBot•1/30/24, 2:43 PM

Thanks again for the additional information. I was able to reproduce the 1031 error by leaving a wrangler dev --remotewrangler dev --remote session open for a couple hours then sending a request to it. Speaking to a colleague, this is a known issue and something we have a plan for addressing as part of the work we're doing to rearchitect

wrangler dev

wrangler dev

. Do you have a rough idea of how long you have to wait for this error to occur? I suspect it's less than 2 hours? For now, I know this isn't ideal, but I'd encourage you to restart wrangler dev --remotewrangler dev --remote when this happens, or consider using the fully-local

wrangler dev

wrangler dev

(without the --remote--remote flag).

I think the CODE_MOVEDCODE_MOVED errors are unrelated to the 13011301 errors, and can be safely ignored.

MityaOP•2/1/24, 3:07 PM

Hey, thanks for the update and glad to know the devs understand the issue and have a fix in mind. Yes it's definitely well south of 2 hours before the issue appears.

Actually I can't run locally without the --remote--remote flag, due to another issue I've raised and been discussing with one of your colleagues hyperdriveHeroku Postgres SSL Errors. It's a really weird one; basically, my Worker won't connect to my DB. (The thread says Heroku DB, but I've since tried it with an Azure-hosted DB and same issue.) The DB is totally open, i.e. no IP restrictions, and other (non-Worker) apps running on my machine, as well as installed DB clients, can connect to the DB fine. The only way round this is to use --remote--remote with a Hyperdrive.

Never a dull moment!

MityaOP•2/2/24, 7:06 PM

Here's another one for you (sorry!), just in case this one is unknown. This happens a lot, seemingly after code changes. It gets stuck on this, and I have to restart the worker.

MMrBBot Thanks again for the additional information. I was able to reproduce the 1031 er...

Shihmin Lee•3/2/24, 1:57 PM

Hi. I kept encountering the same issue repeatedly. I had to restart the local workers, but this is getting annoying as I have more workers running locally. Are there any workarounds for this issue?

"CODE_MOVED for unknown code block" bug?

Similar Threads

Similar Threads

Similar Threads