Unable to run big docker image - 2GB size

[topaz chasm]

Is there a limitation on the image size dagger support?

I cannot run an image locally and on a Kubernetes cluster.
The image size is 2.5 GB.

Local pc trace:
https://v3.dagger.cloud/KubiyaAI/traces/ab9f636a4a41a64cb0a5eb789c6843b6

Kubernetes Trace:
https://v3.dagger.cloud/KubiyaAI/traces/00c4af65516c3e6165bdfce00267a028#debc05da1203a73a

[zealous galleon]

Hi @topaz chasm 👋 Image sizes like that shouldn't be a problem.

I was able to use the image in your trace with no issues. I initiated a new module using the Python SDK, and used the default container_echo generated function with your image instead of alpine:latest:

returning Dagger container type

kubiya-repro ➤ dagger call container-echo --string-arg hello

_type: Container
defaultArgs: []
entrypoint:
    - /bin/sh
    - -c
    - /entry.sh
mounts: []
platform: linux/amd64
user: appuser
workdir: /tmp/workspace

getting the stdout

kubiya-repro ➤ dagger call container-echo --string-arg hello stdout

✔ connect 0.5s
✔ load module 0.9s

✔ kubiyaRepro: KubiyaRepro! 1.3s
$ .containerEcho(stringArg: "hello"): Container! 1.9s CACHED
✔ .stdout: String! 0.0s

✔ parsing command line arguments 0.0s

hello

Are you trying to run as a dagger service, by any chance?
Could be something around your entrypoint script and the way dagger handles entrypoints and services. There were some changes between 0.14 and 0.15 versions.

[topaz chasm]

Im working with dagger service and with version 0.11

[topaz chasm]

We are now able to run the image, but it's taking significantly longer on the Dagger service infrastructure inside the Kubernetes cluster.

It takes around 10 minutes on the cluster, compared to just 2 minutes locally.

What steps can we take to improve this performance?

[zealous galleon]

Yes that container has entrypoint defined. In 0.12 with_exec stopped using the entrypoint by default. In 0.15 as_service stopped using it by default as well.

In 0.11, you're seeing the old behavior of Dagger running the Container's entrypoint script every time.

If you look at the entrypoint script: /entry.sh
You'll see the source of the namespace error messages.

[topaz chasm]

We are now able to run the image, but it's taking significantly longer on the Dagger service infrastructure inside the Kubernetes cluster.

It takes around 10 minutes on the cluster, compared to just 2 minutes locally.

What steps can we take to improve this performance?

[zealous galleon]

Can share Dagger Cloud run links we can look at? 🙏

@rugged hollow @quasi arrow have you seen behavior like this in k8s 👆

[quasi arrow]

Does it take longer to download any image or specifically dagger's engine image? What happens when you try to download ubuntu:latest for example?

[rugged hollow]

This usually happens when there is a difference in system resources, or network configuration.

Are you able to share a single command that we can run @topaz chasm to reproduce it on our end?

[zealous galleon]

Might be here: https://v3.dagger.cloud/KubiyaAI

[rugged hollow]

I can see a lot of local traces, and no CI traces, but I can't see the one command to reproduce it.

[topaz chasm]

it seems like dagger v0.11 is not rendered in the new UI, and for some reason the I cant move to the v2 UI

if you can please enable it for me, I will be able to look for the exact trace for you

[topaz chasm]

@zealous galleon

[zealous galleon]

how do you invoke dagger?
dagger run go run main.go style or dagger call...

[topaz chasm]

using go sdk

[zealous galleon]

Want to pair on discord?

[topaz chasm]

I'm currently tied up with something, but I’d be happy to hop on a call later
In the meantime, Im unable to connect to the traces UI v2, can you help me out with this?

[zealous galleon]

I'm looking into how to view traces for 0.11.6 or workarounds.

[zealous galleon]

@topaz chasm
This should get you there
https://dagger.cloud/KubiyaAI/traces/6d3391f6b22d044e82cd3133a6b46761?v=traces note the lack of leading v3. and the addition of ?v=traces to end. After that you should be able to see them all at:
https://dagger.cloud/KubiyaAI/traces

We're working to ensure the v2 button works properly. Sorry about that, thanks for reporting!!
https://github.com/dagger/dagger.io/pull/4218

[zealous galleon]

Let me know if your slowness issues cleared up 🙏

[topaz chasm]

Hi Jeremy

It hasn't, we are working on decreasing the image size ..

On another matter - creating a volume with files takes about 6 seconds each - is it supposed to be this way?

@zealous galleon

[topaz chasm]

@zealous galleon ??

[zealous galleon]

@sleek bane can you help drive this investigation? 👆

@topaz chasm you're still on dagger 0.11.6, correct?

[hidden musk]

6 seconds for a cert is definitely not right... does your image have 1 layer that's particularly large?

[sleek bane]

@hidden musk looks like very small files in the screenshot its referencing 1445 bytes

@topaz chasm is there a reason you are still on 11.6? It would be great to check to see if these issues are resolved on the latest version of Dagger.