Is it when you're making a rollout? Are your containers setup to handle the `SIGTERM` signal from th

Kkian Is it when you're making a rollout? Are your containers setup to handle the `SIG...

K

Kirk Ouimet•9/30/25, 5:25 PM

OK @kian I figured it out after lots of trial and error. Something happened on Cloudflares end that is breaking

POST

POST

requests from the Workers to the Containers on the first try. It causes this error:

Container suddenly disconnected, try again

Container suddenly disconnected, try again

Container suddenly disconnected, try again

Container suddenly disconnected, try again

So, I implemented retry logic for POST requests that waits 1 second and retries the POST request again and again. It usually works within the second or third request.

GET requests work perfectly every time

My POST retry flow works for now - hope this is helpful

K

Kirk Ouimet•9/30/25, 6:01 PM

And some notes from Claude to the containers team

Screenshot_2025-09-30_at_12.01.22_PM.png

P

Phatso•9/30/25, 7:20 PM

so in the base containers config they simply reference a dockerfile like this: https://github.com/cloudflare/templates/blob/main/containers-template/wrangler.jsonc#L13

that's not ideal for my project because:
1. I want to build the containers myself
- (I need to pass ARGs and then minify them)
2. I want to keep my

docker/

docker/

docker/

docker/ stuff separate of my

web/

web/

web/

web/ stuff

I made this project a few days after Containers came out. Are there any new built-in mechanisms to help me handle this?

Additionally, I need to periodically update my images in the container repo. So far I've been using a Github Action that builds and pushes the new versions using the wrangler CLI
, however this seems like it.. breaks(?) my images? Eventually, after a few updates, when a container boots up it says it can't find the image.

To fix this, I've just had to manually build + minify + push them on my machine.

Any advice would be appreciated!

KKirk Ouimet OK @kian I figured it out after lots of trial and error. Something happened on C...

P

Pranjal•10/1/25, 5:27 AM

I’m experiencing the same issue — everything works as expected in the development environment, but in production I’m getting a

Container suddenly disconnected, try again

Container suddenly disconnected, try again

Container suddenly disconnected, try again

Container suddenly disconnected, try again error with status code 500, even though my container is actually responding with status code 409 or 201. The logs show that the service inside the container returns the correct status code. This issue occurs specifically on POST requests.
Container App ID: a037b631-8b38-4ea3-9ca6-1c094184df54.

A

arnar•10/1/25, 3:30 PM

I'm getting a lot of issues like this

Failed to start container: start() cannot be called on a container that is already running.

Failed to start container: start() cannot be called on a container that is already running.

Failed to start container: start() cannot be called on a container that is already running.

Failed to start container: start() cannot be called on a container that is already running.
I'm using the

getRandom

getRandom

getRandom

getRandom method of selecting a container so some requests might be hitting the same cold container when running requests. But I'm never calling

start()

start()

start()

start() myself, so it seems like an issue in the internal code of the

cloudflare:containers

cloudflare:containers

cloudflare:containers

cloudflare:containers package.

I'm also getting this error fairly frequently

There is no Container instance available at this time.
This is likely because you have reached your max concurrent instance count (set in wrangler config) or are you currently provisioning the Container.
If you are deploying your Container for the first time, check your dashboard to see provisioning status, this may take a few minutes.

There is no Container instance available at this time.
This is likely because you have reached your max concurrent instance count (set in wrangler config) or are you currently provisioning the Container.
If you are deploying your Container for the first time, check your dashboard to see provisioning status, this may take a few minutes.

There is no Container instance available at this time.
This is likely because you have reached your max concurrent instance count (set in wrangler config) or are you currently provisioning the Container.
If you are deploying your Container for the first time, check your dashboard to see provisioning status, this may take a few minutes.

There is no Container instance available at this time.
This is likely because you have reached your max concurrent instance count (set in wrangler config) or are you currently provisioning the Container.
If you are deploying your Container for the first time, check your dashboard to see provisioning status, this may take a few minutes.

Even though there's no rollout going on and I have plenty of available machines (either stopped or started).

Any specific solution to this or just a symptom of it still being in beta?

D

dr.dj•10/1/25, 5:25 PM

Will Containers have access to D1's like Workers?

M

Mike Nomitch•10/1/25, 8:57 PM

Both are planned.

First step -> We're working on generic requests from container to worker on specified hostnames. IE:

foo.cf

foo.cf

foo.cf

foo.cf in the container calls to

handleFoo

handleFoo

handleFoo

handleFoo. - That will open up a bunch of options, and then from there we'll build out more official integrations with D1 and hyperdrive.

R2 also definitely coming, but not as sure on timeline yet. The step above would allow you to manually set it up, but the ideal interface is just a mounted volume IMO.

M

Mike Nomitch•10/1/25, 8:58 PM

Also...

Larger Container Instances are live! - https://developers.cloudflare.com/changelog/2025-10-01-new-container-instance-types/

Get up to 4 vCPU, 12 GiB memory, and 20 GB Disk.

M

Mike Nomitch•10/1/25, 9:02 PM

( next up: pricing update and custom instance types)

I

iddogino•10/2/25, 12:10 AM

@Mike Nomitch | Workers PM love the larger instance types! Using the latest Wrangler it's complaining about the instane types not being valid (still looks for the valu to be

["dev","basic","standard"]

["dev","basic","standard"]

["dev","basic","standard"]

["dev","basic","standard"]) -- is there a canary or something I should be using?

Iiddogino @Mike Nomitch | Workers PM love the larger instance types! Using the latest Wran...

C

Chaika•10/2/25, 12:24 AM

There's a PR for it https://github.com/cloudflare/workers-sdk/pull/10844 and a linked preview but looks like it'll be out pretty soon

CChaika There's a PR for it https://github.com/cloudflare/workers-sdk/pull/10844 and a l...

I

iddogino•10/2/25, 1:13 AM

Awesome, thanks

D

Didac•10/2/25, 7:19 AM

hey guys, I thought I was going crazy but from what I can gather, the Container suddenly disconnected problem is a cloudflare issue? and the only way to workaround it is just to implement a retry logic?

D

Didac•10/2/25, 7:42 AM

i can confirm the retry logic works but it's crazy to have to do that :NotLikeThis:

DDidac hey guys, I thought I was going crazy but from what I can gather, the Container ...

K

kianOP•10/2/25, 8:41 AM

We're already actively looking into it & have a possible fix on the way.

MMike Nomitch Both are planned. First step -> We're working on generic requests from containe...

K

Karthik Ramani•10/2/25, 9:06 AM

so people in the pro plan can use standard-4 4 12 GiB 20 GB container right? we are trying the vibe SDK which is recently launched, in that it says you need an enterprise plan, please clarify? will they update it?

D

Didac•10/2/25, 10:15 AM

i have a project that i run with react-router dev, is it possible to run containers at the same time with this on the same project, same wrangler config? i tried and i get workerd/jsg/util.c++:320: error: e = kj/async-io-unix.c++:1509: failed: broken.constructorFailed; jsg-internal.Error: connect(): No such file or directory; address = unix:///var/run/docker.sock
stack: 102b17597 102b1540f 102b15a50 10091350c 100c0b747 100c0d05b 100c12d27 1013e96c0 1013e39b4 100931568 100f00c1b 100948f7f 1009232d3 100949e13 10125e57c 101260a7f 100f032cb 1008dfb38 10107302c 100f86538; sentryErrorContext = jsgInternalError; wdErrId = u6uc42g0bpojgvbp9g8mt5h2.

DDidac i have a project that i run with react-router dev, is it possible to run contain...

D

Didac•10/2/25, 10:16 AM

but the container on its own worker project works fine. when running alongside react-router dev, i get that error. docker image does seem to build though but i don't get the "container image(s) ready"

D

Didac•10/2/25, 12:36 PM

Works only with wrangler dev.. it’s alright

T

tao•10/2/25, 1:17 PM

whenever I deploy a new version, there is a brief downtime while containers instance restart. I’d like to reduce or eliminate that downtime.Any good suggestions?

K

kianOP•10/2/25, 1:24 PM

There shouldn't really be any downtime with multiple instances

S

Sajad•10/2/25, 1:26 PM

hi

S

Sajad•10/2/25, 1:27 PM

Why is the site no longer working?
https://tt886399.com

T

tao•10/2/25, 1:27 PM

is it possible to configure multiple instances for my container ?

T

tao•10/2/25, 1:29 PM

rn i only have single instace,since we are still in the staging phase and the instance often goes inactive.

K

kianOP•10/2/25, 1:46 PM

https://developers.cloudflare.com/containers/platform-details/scaling-and-routing/#scaling-container-instances-with-get - you can set a

max_instances

max_instances

max_instances

max_instances in Wrangler but otherwise it is similar to Durable Objects where each Container instance has it's own name that you provide.

K

kianOP•10/2/25, 1:51 PM

Basically if you have a single instance (i.e taken to mean you never want two instances of your code running at once) then there is the inevitable blip where we take down the instance to restart it with your new image.

K

kianOP•10/2/25, 1:51 PM

If you're happy to have multiple instances at once then you can load balance between them (we have a function in the Containers lib to make that easy)

F

fforres•10/2/25, 2:58 PM

Hey folks (maybe @kian or @Mike Nomitch )
I remember a call some time ago where we discussed some caveats whith locations where containers would actually be available/served from v/s their durable objects. (DOs where not co-located with a container. I think that's still the case, and ok by us)

So mostly my question is on the locations where the containers are served from. I think I remember this was written down somewhere? (This was pre public beta , so also could've been some stuff written mid-meeting) But cannot seem to find this.

For context, my company's main product relies on a sync engine. The engine being close to our users is kind of important (not as much as the engine being close to DB, but still)

Back then the locations of containers it was a bit of an issue for us, wanted to reevaluate if we could move out of our current provider.

Ffforres Hey folks (maybe @kian or @Mike Nomitch ) I remember a call some time ago where...

H

Hello, I’m Allie!•10/2/25, 3:20 PM

?pings

S

SuperHelpflare•10/2/25, 3:20 PM

Please do not ping community members for non-moderation reasons. Doing so will not solve your issue faster and will make people less likely to want to help you.

S

Salem•10/2/25, 3:32 PM

Is anyone else experiencing wrangler deploy hanging on one Layer?

SSalem Is anyone else experiencing wrangler deploy hanging on one Layer?

S

Sean•10/2/25, 4:31 PM

Yes, getting this too. Also just complete deployment failures:

╭ Deploy a container application deploy changes to your application
│

✘ [ERROR] Unauthorized


If you think this is a bug then please create an issue at https://github.com/cloudflare/workers-sdk/issues/new/choose

╭ Deploy a container application deploy changes to your application
│

✘ [ERROR] Unauthorized


If you think this is a bug then please create an issue at https://github.com/cloudflare/workers-sdk/issues/new/choose

╭ Deploy a container application deploy changes to your application
│

✘ [ERROR] Unauthorized


If you think this is a bug then please create an issue at https://github.com/cloudflare/workers-sdk/issues/new/choose

╭ Deploy a container application deploy changes to your application
│

✘ [ERROR] Unauthorized


If you think this is a bug then please create an issue at https://github.com/cloudflare/workers-sdk/issues/new/choose

S

Sean•10/2/25, 4:32 PM

Also getting the sudden disconnection issue mentioned above

K

kianOP•10/2/25, 4:33 PM

Does

npx wrangler whoami

npx wrangler whoami

npx wrangler whoami

npx wrangler whoami show the Containers scope?

S

Sean•10/2/25, 4:35 PM

Yes, that was the first time I've seen this issue. I tried re-running the deploy command and it's hanging at one layer.

Screenshot_2025-10-02_at_11.35.50_AM.png

S

Salem•10/2/25, 4:37 PM

I just closed down the terminal and reopened it and it worked eventually after a few tries

Kkian We're already actively looking into it & have a possible fix on the way.

A

arnar•10/3/25, 11:54 AM

Will you notify in this channel when this potential fix has been released?
It seems from our testing that this is the only issue we have blocking us from starting to route some small part of our production traffic to CF Containers

Aarnar Will you notify in this channel when this potential fix has been released? It se...

S

Sean•10/3/25, 4:24 PM

Second this, seems to be the only issue blocking us from deploying

P

piemaking•10/3/25, 6:44 PM

Trying to setup a container from within a NextJs obviously via Opennext.

It’s failing on loading “Cloudflare:workers” from the containers library,

D

dr.dj•10/4/25, 3:45 PM

Hi all, is there an easy way to tail logs inside your container, ala

npm wrangler tail some.container.id

npm wrangler tail some.container.id

npm wrangler tail some.container.id

npm wrangler tail some.container.id?

A

Alex Dahl•10/5/25, 9:25 AM

Containers look great! I'm currently running a bunch of machines on fly.io for my workload. (fast api + duckdb) The only feature missing from containers seem to be the lack of persistent disk, which I use for the duckdb file. I think I can work around that with R2 and lifespan in fast api though!

Can we do an R2 binding for containers or should I set up a typical access/secret key thing?

Edit: pre-emptively answered here! containers-beta

Cloudflare Docs

Frequently Asked Questions

Frequently Asked Questions:

K

kingkm70•10/5/25, 2:31 PM

Hello - I have a couple of issues with containers.

#1) My container isn't logging anything. I can see a call that passed to container, but nothing shows up in the container logger.

For example:

I have a simple endpoint that passes a ping to the container...

./scripts/ping.sh
- Target URL: https://<cfworker>.workers.dev
{
"success": true
}

This I see in my CF Worker log...
2025-10-05 14:13:01:654 UTC
GET /ping
Error checking 8080: The container is not listening in the TCP address 10.0.0.1:8080

But the:

{
"success": true
}

is returned from my container.

#2) My container is trashing. Meaning I'm seeing it being shutdown every couple of minutes.

I have it set for sleepAfter = "60m"; -- so it shouldn't be going down randomly...

Any insights would be greatly appreciated.

F

FiveGuys•10/5/25, 5:58 PM

Was in the private BETA and now getting

The image registry does not exist

The image registry does not exist

The image registry does not exist

The image registry does not exist. User account is linked to profile.

Is it when you're making a rollout? Are your containers setup to handle the `SIGTERM` signal from th

Similar Threads