Hi all, we use Claude Agent SDK for all of our agent workflows and have just figured out how to get

M

Martijn•10/17/25, 7:32 PM

Hi,

With the new sandbox that just came out are there plans to make it possible to deploy to a custom domain? For example I would like to deploy every PR to

preview.domain.com

preview.domain.com

preview.domain.com

preview.domain.com and have the PR number be either a prefix of the domain or an url so like

<pr>.preview.domain.com

<pr>.preview.domain.com

<pr>.preview.domain.com

<pr>.preview.domain.com or

preview.domain.com/<pr>

preview.domain.com/<pr>

preview.domain.com/<pr>

preview.domain.com/<pr>

K

Kris•10/18/25, 2:21 AM

Just in case someone else hits this in the future and searches: Bad things happen if there's a mismatch between the version of the

@cloudflare/sandbox

@cloudflare/sandbox

@cloudflare/sandbox

@cloudflare/sandbox package and the

docker.io/cloudflare/sandbox

docker.io/cloudflare/sandbox

docker.io/cloudflare/sandbox

docker.io/cloudflare/sandbox Docker image. Was getting only a very generic

SandboxError

SandboxError

SandboxError

SandboxError with no other details which made it hard to track down.

Ttao How can I differentiate between production and staging environments when deployi...

P

pandalambda2•10/18/25, 6:24 AM

at this point its easier to use sam's alchemy or pulumi, think of wrangler as the fisher price toy, great for starting but eventually you'll be wrapping in your own scripts or shelling out, you might as well adopt a proper iac solution that has native concepts of environments (stacks in pulumi)

V

varasha_•10/18/25, 9:10 AM

Hello guys, can im use import http2 from 'http2' on cloud falre container?

J

John Spurlock•10/18/25, 3:40 PM

Looking forward to hearing about new pricing! As a baseline, here are the currently-documented costs for each instance type of running a single container 24/7 for 30 days (also the per hour cost) for easier ballpark comparisons. Container costs only, ignores egress and the cost of the DO keeping it alive. Let me know if my math is off

Hhudge Hi all, we use Claude Agent SDK for all of our agent workflows and have just fig...

K

Kris•10/18/25, 10:25 PM

I gave it a try as well, using the new @cloudflare/sandbox package. I found Claude Code to slow for my case, which is basically a one-shot generation of new code (rather than needing to modify existing code). For this, I found generating the code outside the container (using the Claude API directly) and just building it inside the container was much faster.

J

js•10/19/25, 2:03 PM

is it possible to do sticky routing for containers? i have this thing where each tenant needs to start containers with their own tenant id for everything to work, but i don't want to create one statically configured individual container binding per container. is this possible somehow? or is dynamic binding a thing? also i realize that i need to pass custom env to the containers, is that possible runtime?

Jjs is it possible to do sticky routing for containers? i have this thing where each...

K

kian•10/19/25, 6:50 PM

What is actually different between the Containers per tenant?

& yes you can pass environment variables to the start function

Kkian What is actually different between the Containers per tenant? & yes you can pas...

J

js•10/19/25, 7:18 PM

Those env vars sre different. But i think getContainer with a naming pattern could be the say to go

D

darkpool•10/20/25, 7:44 PM

is there anything we can do with wrangler's dev registry to bypass dockerizing the container each time we want to run it locally? I am new to docker and containerization in general so not sure if I'm missing something key but waiting for container to build during dev is a slog in a monorepo. When I'm just working on the container I can just run it directly with tsx but Id like a way to just tell the CF container class to look for a local instance running (simmilar to how wrangler finds other workers exposing RPC entrypoints) EDIT: just pushing and pulling from registry is what I needed to do

A

Alex Dahl•10/20/25, 8:30 PM

I got around to testing download speed from R2 into a container. A bit disappointed, roughly the same speed as just

wget

wget

on from my home network. :sad_frog: (using a public R2 bucket for the test)

data:
lite:
{"duration_seconds":8.887570779,"file_size_mb":104.54691410064697,"speed_mbps":94.10617744743091}

standard-1
{"duration_seconds":2.787745816,"file_size_mb":104.54691410064697,"speed_mbps":300.0184981015413}
{"duration_seconds":2.085034671,"file_size_mb":104.54691410064697,"speed_mbps":401.13256841146114}
{"duration_seconds":2.383474448,"file_size_mb":104.54691410064697,"speed_mbps":350.90592790159246}

standard-2
{"duration_seconds":3.066984332,"file_size_mb":104.54691410064697,"speed_mbps":272.7028319247298}
{"duration_seconds":2.242448686,"file_size_mb":104.54691410064697,"speed_mbps":372.97411442536605}
{"duration_seconds":2.835909462,"file_size_mb":104.54691410064697,"speed_mbps":294.9231363032759}

standard-3
{"duration_seconds":2.260594997,"file_size_mb":104.54691410064697,"speed_mbps":369.98016624610614}
{"duration_seconds":2.29764062,"file_size_mb":104.54691410064697,"speed_mbps":364.014853117097}
{"duration_seconds":2.292594704,"file_size_mb":104.54691410064697,"speed_mbps":364.81603632160176}

Z

Zentinela•10/22/25, 8:28 AM

Hi everyone!

I upgraded to Workers Paid plan ($5/month) 13+ hours ago but Containers are still not enabled for my Durable Object.

Account: Hello@bechallenge.io
Worker: vibesdk-production.cognicraft.workers.dev
Error: "Container is not enabled for this durable object class"

Workers Paid active (payment confirmed: $5 USD/month)

@cloudflare/sandbox 0.1.3 installed

Waited 13+ hours

Redeployed multiple times

Dashboard shows "You have no Containers"

Can someone help enable Containers for my account?

ZZentinela Hi everyone! 👋 I upgraded to Workers Paid plan ($5/month) 13+ hours ago but Co...

K

kian•10/22/25, 2:36 PM

"Container is not enabled for this durable object class" is likely an accurate error - what's your Wrangler config?

N

nicoritschel•10/22/25, 2:45 PM

I seem to have inconsistency in the state of my container. If I click that instance marked stopped it seems nothing is running as per metrics. I expect that something is running currently as it was triggered from a workflow.

N

nicoritschel•10/22/25, 2:46 PM

The meter also shows 1 live instance; maybe it's not booting?

N

nicoritschel•10/22/25, 2:51 PM

Ah was just intermittent, redeploying cycled it

K

kian•10/22/25, 2:54 PM

There's a little bug we're currently fixing where live instances represents your Durable Object that has access to the Container being active even if your Container itself isn't (i.e DOs hang around for 30s after their last request or you were doing non-Containers API calls)

K

kian•10/22/25, 2:54 PM

That's just a dash / health reporting UI quirk, your Container shouldn't be running & we're not billing you for it.

T

Thibault•10/22/25, 9:03 PM

What would be the best way to keep a container up and running 24/7 ?

N

nicoritschel•10/22/25, 10:13 PM

You could probably use an alarm on the attached durable object and keep waking it

F

fare•10/23/25, 1:28 AM

js import { Container, getRandom } from "@cloudflare/containers";

interface Env {
  API: any; // Container binding
  API_KEY: string;
  CLEANUP_TEMP_FILES: string;
  LOG_LEVEL: string;
  PORT: string;
  R2_ACCESS_KEY_ID: string;
  R2_BUCKET_NAME: string;
  R2_ENDPOINT_URL: string;
  R2_PUBLIC_URL: string;
  R2_SECRET_ACCESS_KEY: string;
}

export class PowerpointApi extends Container {
  declare env: Env;
  defaultPort = 8000;
  sleepAfter = "2m";
  envVars: Record<string, string> = {
    API_KEY: this.env.API_KEY,
    CLEANUP_TEMP_FILES: this.env.CLEANUP_TEMP_FILES,
    LOG_LEVEL: this.env.LOG_LEVEL,
    PORT: this.env.PORT,
    R2_ACCESS_KEY_ID: this.env.R2_ACCESS_KEY_ID,
    R2_BUCKET_NAME: this.env.R2_BUCKET_NAME,
    R2_ENDPOINT_URL: this.env.R2_ENDPOINT_URL,
    R2_PUBLIC_URL: this.env.R2_PUBLIC_URL,
    R2_SECRET_ACCESS_KEY: this.env.R2_SECRET_ACCESS_KEY,
  };

  override onStart() {
    console.log("Container successfully started");
  }

  override onStop() {
    console.log("Container successfully shut down");
  }

  override onError(error: unknown) {
    console.log("Container error:", error);
  }
}


export default {
  async fetch(request: Request, env: Env): Promise<Response> {
    const container = await getRandom(env.API, 25);
    return await container.fetch(request);
  },
} satisfies ExportedHandler<Env>;

js import { Container, getRandom } from "@cloudflare/containers";

interface Env {
  API: any; // Container binding
  API_KEY: string;
  CLEANUP_TEMP_FILES: string;
  LOG_LEVEL: string;
  PORT: string;
  R2_ACCESS_KEY_ID: string;
  R2_BUCKET_NAME: string;
  R2_ENDPOINT_URL: string;
  R2_PUBLIC_URL: string;
  R2_SECRET_ACCESS_KEY: string;
}

export class PowerpointApi extends Container {
  declare env: Env;
  defaultPort = 8000;
  sleepAfter = "2m";
  envVars: Record<string, string> = {
    API_KEY: this.env.API_KEY,
    CLEANUP_TEMP_FILES: this.env.CLEANUP_TEMP_FILES,
    LOG_LEVEL: this.env.LOG_LEVEL,
    PORT: this.env.PORT,
    R2_ACCESS_KEY_ID: this.env.R2_ACCESS_KEY_ID,
    R2_BUCKET_NAME: this.env.R2_BUCKET_NAME,
    R2_ENDPOINT_URL: this.env.R2_ENDPOINT_URL,
    R2_PUBLIC_URL: this.env.R2_PUBLIC_URL,
    R2_SECRET_ACCESS_KEY: this.env.R2_SECRET_ACCESS_KEY,
  };

  override onStart() {
    console.log("Container successfully started");
  }

  override onStop() {
    console.log("Container successfully shut down");
  }

  override onError(error: unknown) {
    console.log("Container error:", error);
  }
}


export default {
  async fetch(request: Request, env: Env): Promise<Response> {
    const container = await getRandom(env.API, 25);
    return await container.fetch(request);
  },
} satisfies ExportedHandler<Env>;

js import { Container, getRandom } from "@cloudflare/containers";

interface Env {
  API: any; // Container binding
  API_KEY: string;
  CLEANUP_TEMP_FILES: string;
  LOG_LEVEL: string;
  PORT: string;
  R2_ACCESS_KEY_ID: string;
  R2_BUCKET_NAME: string;
  R2_ENDPOINT_URL: string;
  R2_PUBLIC_URL: string;
  R2_SECRET_ACCESS_KEY: string;
}

export class PowerpointApi extends Container {
  declare env: Env;
  defaultPort = 8000;
  sleepAfter = "2m";
  envVars: Record<string, string> = {
    API_KEY: this.env.API_KEY,
    CLEANUP_TEMP_FILES: this.env.CLEANUP_TEMP_FILES,
    LOG_LEVEL: this.env.LOG_LEVEL,
    PORT: this.env.PORT,
    R2_ACCESS_KEY_ID: this.env.R2_ACCESS_KEY_ID,
    R2_BUCKET_NAME: this.env.R2_BUCKET_NAME,
    R2_ENDPOINT_URL: this.env.R2_ENDPOINT_URL,
    R2_PUBLIC_URL: this.env.R2_PUBLIC_URL,
    R2_SECRET_ACCESS_KEY: this.env.R2_SECRET_ACCESS_KEY,
  };

  override onStart() {
    console.log("Container successfully started");
  }

  override onStop() {
    console.log("Container successfully shut down");
  }

  override onError(error: unknown) {
    console.log("Container error:", error);
  }
}


export default {
  async fetch(request: Request, env: Env): Promise<Response> {
    const container = await getRandom(env.API, 25);
    return await container.fetch(request);
  },
} satisfies ExportedHandler<Env>;

js import { Container, getRandom } from "@cloudflare/containers";

interface Env {
  API: any; // Container binding
  API_KEY: string;
  CLEANUP_TEMP_FILES: string;
  LOG_LEVEL: string;
  PORT: string;
  R2_ACCESS_KEY_ID: string;
  R2_BUCKET_NAME: string;
  R2_ENDPOINT_URL: string;
  R2_PUBLIC_URL: string;
  R2_SECRET_ACCESS_KEY: string;
}

export class PowerpointApi extends Container {
  declare env: Env;
  defaultPort = 8000;
  sleepAfter = "2m";
  envVars: Record<string, string> = {
    API_KEY: this.env.API_KEY,
    CLEANUP_TEMP_FILES: this.env.CLEANUP_TEMP_FILES,
    LOG_LEVEL: this.env.LOG_LEVEL,
    PORT: this.env.PORT,
    R2_ACCESS_KEY_ID: this.env.R2_ACCESS_KEY_ID,
    R2_BUCKET_NAME: this.env.R2_BUCKET_NAME,
    R2_ENDPOINT_URL: this.env.R2_ENDPOINT_URL,
    R2_PUBLIC_URL: this.env.R2_PUBLIC_URL,
    R2_SECRET_ACCESS_KEY: this.env.R2_SECRET_ACCESS_KEY,
  };

  override onStart() {
    console.log("Container successfully started");
  }

  override onStop() {
    console.log("Container successfully shut down");
  }

  override onError(error: unknown) {
    console.log("Container error:", error);
  }
}


export default {
  async fetch(request: Request, env: Env): Promise<Response> {
    const container = await getRandom(env.API, 25);
    return await container.fetch(request);
  },
} satisfies ExportedHandler<Env>;

There is no container instance that can be provided to this Durable Object, try again later - we are also getting this error a lot I am adding the config below.

image = "./Dockerfile"
max_instances = 25
instance_type = "standard-4"

image = "./Dockerfile"
max_instances = 25
instance_type = "standard-4"

image = "./Dockerfile"
max_instances = 25
instance_type = "standard-4"

image = "./Dockerfile"
max_instances = 25
instance_type = "standard-4"

Could you hint us that where are we going wrong for these two places?

Ffare ```js import { Container, getRandom } from "@cloudflare/containers"; interface ...

K

kingkm70•10/23/25, 2:28 AM

From my understanding cloudflare uses alarm to validate/keep your container up.

Kkingkm70 From my understanding cloudflare uses alarm to validate/keep your container up.

F

fare•10/23/25, 2:44 AM

oh just got a little worried as I see a lot of alarm error in my dashboard

F

fare•10/23/25, 2:45 AM

But I cant still get my head around the container instance. Is there now way to increase containers? I have already givem the maximum thats available is my understanding wrong somewhere?

K

kingkm70•10/23/25, 2:48 AM

I don't now about that your random - to load balance? Seems like you should be letting cloudflare handler distribution.

Nnicoritschel You could probably use an alarm on the attached durable object and keep waking i...

T

Thibault•10/23/25, 7:58 AM

This is what i'm trying

  const container = getContainer(env.MY_CONTAINER);
      const containerState = await container.getState();
      if (containerState.status === "stopped") {
        await container.startAndWaitForPorts({
          startOptions: {
            envVars: {
              OPENSTATUS_KEY: env.OPENSTATUS_KEY,
            },
          },
        });
      } else {
        await container.renewActivityTimeout();
      }

  const container = getContainer(env.MY_CONTAINER);
      const containerState = await container.getState();
      if (containerState.status === "stopped") {
        await container.startAndWaitForPorts({
          startOptions: {
            envVars: {
              OPENSTATUS_KEY: env.OPENSTATUS_KEY,
            },
          },
        });
      } else {
        await container.renewActivityTimeout();
      }

  const container = getContainer(env.MY_CONTAINER);
      const containerState = await container.getState();
      if (containerState.status === "stopped") {
        await container.startAndWaitForPorts({
          startOptions: {
            envVars: {
              OPENSTATUS_KEY: env.OPENSTATUS_KEY,
            },
          },
        });
      } else {
        await container.renewActivityTimeout();
      }

  const container = getContainer(env.MY_CONTAINER);
      const containerState = await container.getState();
      if (containerState.status === "stopped") {
        await container.startAndWaitForPorts({
          startOptions: {
            envVars: {
              OPENSTATUS_KEY: env.OPENSTATUS_KEY,
            },
          },
        });
      } else {
        await container.renewActivityTimeout();
      }

But there's some error in the worker and wall time is blocked

D

dhrubo•10/23/25, 6:43 PM

Can I deploy java spring boot docker images on the container platform ?

Ffare ```js import { Container, getRandom } from "@cloudflare/containers"; interface ...

F

fare•10/24/25, 4:07 AM

Hi team, Following up on this is there an update?

Ffare Hi team, Following up on this is there an update?

1

1984 Ford Laser•10/24/25, 4:32 AM

Are you exceeding the max available instances/resources limits?

11984 Ford Laser Are you exceeding the max available instances/resources limits?

F

fare•10/24/25, 4:46 AM

Yes, it says
There is no container instance that can be provided to this Durable Object, try again later

and when I try to increase instances in wrangler.toml myself it says

SURPASSED_TOTAL_LIMITS
reason This application requires more than the allocated quota of vcpus in
this account to run all its requested instances of count 40.
used_vcpu 0
remaining_vcpu 100
requested_vcpu 160

1

1984 Ford Laser•10/24/25, 4:46 AM

>requested_vcpu - 160
>remaining_vcpu - 100

That might be an issue lol

11984 Ford Laser >requested_vcpu - 160 >remaining_vcpu - 100 That might be an issue lol

F

fare•10/24/25, 4:47 AM

Yes, but how do I get more instances I just need them.

Our prod isn't scaling an it has become an issue

1

1984 Ford Laser•10/24/25, 4:49 AM

Decrease max instances to stay under the vcpu limit. These limits will be increased over time as the platform matures and kinks are worked out

F

fare•10/24/25, 4:52 AM

this message doesnt show up in production, it shows when i try to deploy the wrangler.toml with more instances
the reason why I am increasing it because we are getting this error

There is no container instance that can be provided to this Durable Object, try again later

There is no container instance that can be provided to this Durable Object, try again later

There is no container instance that can be provided to this Durable Object, try again later

There is no container instance that can be provided to this Durable Object, try again later

1

1984 Ford Laser•10/24/25, 4:59 AM

Change

getRandom(env.API, 25)

getRandom(env.API, 25)

getRandom(env.API, 25)

getRandom(env.API, 25) to a more reasonable value

1

1984 Ford Laser•10/24/25, 4:59 AM

You obviously don't have 25 instances available

1

1984 Ford Laser•10/24/25, 5:00 AM

Like if it randomly routes to container # 25 does it assume that 1-24 are already consumed or something?

11984 Ford Laser You obviously don't have 25 instances available

F

fare•10/24/25, 5:08 AM

It randomly picks one of 25 instances. That is how getRandom works

L

lemmein•10/24/25, 5:09 AM

Hi. I have been tricked by an ad/review on the website www.theresanaiforthat.com into purchasing a year subscription of www.v93ai . The former Site advertised this A I Tool specifically as an "Illegal" Video Generator , meaning basically that it was promoted as an Agent which can be used to generate Adult or Explicit Content. For $82.00 I discovered instantly that it did not! I've emailed the company outlining my dissatisfaction but
it hasn't got through, it just returned to my email. So I'm here asking you, what is my next step & how do I go about getting a refund from these people? I'd appreciate some advice.
Many thanks.

L

lemmein•10/24/25, 5:10 AM

The Agent is called v03ai.

N

niall•10/24/25, 1:49 PM

since ~2200-0000UTC last night, still ongoing, we've started seeing a whole bunch of CF internal errors thrown by our containers when trying to communicate with DOs

e.g.

internal error; reference = 0sp6nv1fqj06ub0q9b0qbfjq

internal error; reference = 0sp6nv1fqj06ub0q9b0qbfjq

internal error; reference = 0sp6nv1fqj06ub0q9b0qbfjq

internal error; reference = 0sp6nv1fqj06ub0q9b0qbfjq

any ideas? :sweat_heh:

L

Lulu•10/25/25, 7:40 AM

Since DOs are billed separately, I was curious: Is the durable object tied to the container always running as long as the container is? i.e. unable to hibernate and continuously charged

LLulu Since DOs are billed separately, I was curious: Is the durable object tied to th...

K

kian•10/25/25, 1:28 PM

Yes & no - your DO doesn’t necessarily have to be around but currently does influence the lifetime of your Container, I.E sleepAfter in the Containers library starts after DO inactivity

K

kian•10/25/25, 1:29 PM

We’re working on moving lifecycles to more Containers APIs so the DO doesn’t have to be around to do any of that though

K

kian•10/25/25, 1:29 PM

Either way we’re looking at reducing or entirely removing the DO wall-time billing when using Containers: containers-beta

<

<@235148962103951360>•10/25/25, 6:09 PM

Is there a way to run something just before a container goes to sleep/is destroyed? I'd like to collect some files and other info from the container before it is destroyed.
(edit: found the function)

Hi all, we use Claude Agent SDK for all of our agent workflows and have just figured out how to get

Similar Threads

Similar Threads

Similar Threads