2024-03-01T16:08:54.761577365Z [FATAL tini (6)] exec docker failed: No such file or directory Error

voiceofkronos · 2024-03-01T18:03:58.077Z

Hey folks I'm having trouble running my image on Runpod. My image works properly on a normal root access Docker environment but doesn't work on the Runpod template structure. It's my first time on Runpod, can anyone help out?

M

Madiator2011•3/1/24, 6:10 PM

what is the error

A

ashley•3/1/24, 6:11 PM

In the post title:

2024-03-01T16:08:54.761577365Z [FATAL tini (6)] exec docker failed: No such file or directory Error

2024-03-01T16:08:54.761577365Z [FATAL tini (6)] exec docker failed: No such file or directory Error

2024-03-01T16:08:54.761577365Z [FATAL tini (6)] exec docker failed: No such file or directory Error

2024-03-01T16:08:54.761577365Z [FATAL tini (6)] exec docker failed: No such file or directory Error

A

ashley•3/1/24, 6:12 PM

You probably need a CUDA base image for starters

A

ashley•3/1/24, 6:15 PM

FROM runpod/pytorch:2.1.0-py3.10-cuda11.8.0-devel-ubuntu22.04

WORKDIR /app

RUN apt-get update && apt-get install -y \
    git \
    && rm -rf /var/lib/apt/lists/*

COPY requirements.txt /app/

RUN pip install --no-cache-dir diffusers accelerate -r requirements.txt

COPY . /app

ENV MODEL_DEVICE=gpu

COPY --chmod=755 start.sh /start.sh

CMD ["/start.sh"]

FROM runpod/pytorch:2.1.0-py3.10-cuda11.8.0-devel-ubuntu22.04

WORKDIR /app

RUN apt-get update && apt-get install -y \
    git \
    && rm -rf /var/lib/apt/lists/*

COPY requirements.txt /app/

RUN pip install --no-cache-dir diffusers accelerate -r requirements.txt

COPY . /app

ENV MODEL_DEVICE=gpu

COPY --chmod=755 start.sh /start.sh

CMD ["/start.sh"]

FROM runpod/pytorch:2.1.0-py3.10-cuda11.8.0-devel-ubuntu22.04

WORKDIR /app

RUN apt-get update && apt-get install -y \
    git \
    && rm -rf /var/lib/apt/lists/*

COPY requirements.txt /app/

RUN pip install --no-cache-dir diffusers accelerate -r requirements.txt

COPY . /app

ENV MODEL_DEVICE=gpu

COPY --chmod=755 start.sh /start.sh

CMD ["/start.sh"]

FROM runpod/pytorch:2.1.0-py3.10-cuda11.8.0-devel-ubuntu22.04

WORKDIR /app

RUN apt-get update && apt-get install -y \
    git \
    && rm -rf /var/lib/apt/lists/*

COPY requirements.txt /app/

RUN pip install --no-cache-dir diffusers accelerate -r requirements.txt

COPY . /app

ENV MODEL_DEVICE=gpu

COPY --chmod=755 start.sh /start.sh

CMD ["/start.sh"]

Try this.

A

ashley•3/1/24, 6:16 PM

You don't need

EXPOSE

EXPOSE

EXPOSE

EXPOSE, add the ports to your template/pod

V

voiceofkronosOP•3/1/24, 6:57 PM

@ashleyk The cuda base image worked, thank you so much!
2024-03-01T18:56:20.332807898Z /opt/nvidia/nvidia_entrypoint.sh: line 67: exec: docker: not found any idea about this one?

Vvoiceofkronos @ashleyk The cuda base image worked, thank you so much! 2024-03-01T18:56:20.3328...

A

ashley•3/1/24, 6:58 PM

Where do you see this? Are you trying to run it from within your start.sh script?

V

voiceofkronosOP•3/1/24, 6:59 PM

Here it is @ashleyk

Vvoiceofkronos Here it is @ashleyk

A

ashley•3/1/24, 7:00 PM

Did you set a docker start command for your pod or something?

V

voiceofkronosOP•3/1/24, 7:15 PM

@ashleyk , you are right. The default run script was working. I edited it, and now it's up and running.

Last question: I have an endpoint that listens on port 8081. When I send a request, I can't reach the port and can't get results.

How I should configure template config, below on image?

Vvoiceofkronos @ashleyk , you are right. The default run script was working. I edited it, and n...

A

ashley•3/1/24, 7:16 PM

Make your app bind to
```
0.0.0.0
```
0.0.0.0
```
0.0.0.0
```
0.0.0.0
and not
```
127.0.0.1
```
127.0.0.1
```
127.0.0.1
```
127.0.0.1
.
Add HTTP port 8081 to your pod, or alternatively a TCP port if there is a possibility that your request execution time will exceed 100s.

A

ashley•3/1/24, 7:18 PM

I see you already added it as a TCP port. TCP ports require a public IP, so if you are using Community Cloud, you need to ensure that you check the Public IP filter at the top of the page.

A

ashley•3/1/24, 7:18 PM

Then you need to go to the

Connect

Connect

Connect

Connect button TCP port mappings and get the external port mapping and use the external port. 8081 will be the internal port.

V

voiceofkronosOP•3/1/24, 7:18 PM

I'm using secure cloud, when I remove TCP port I'm losing public ip

Vvoiceofkronos I'm using secure cloud, when I remove TCP port I'm losing public ip

A

ashley•3/1/24, 7:20 PM

I always add at least port 22 for SSH access.

V

voiceofkronosOP•3/1/24, 7:27 PM

This is the last config, still can't reach the port. I checked it with telnet. Am I missing something? @ashleyk

V

voiceofkronosOP•3/1/24, 7:34 PM

@ashleyk

A

ashley•3/1/24, 7:36 PM

You probably didn't bind your app to

0.0.0.0

0.0.0.0

0.0.0.0

0.0.0.0 like I said above

A

ashley•3/1/24, 7:37 PM

it can't work if it binds to

127.0.0.1

127.0.0.1

127.0.0.1

127.0.0.1.

V

voiceofkronosOP•3/1/24, 7:39 PM

@ashleyk it's already 0.0.0.0

A

ashley•3/1/24, 7:46 PM

What port did you telnet to?

Aashley What port did you telnet to?

V

voiceofkronosOP•3/1/24, 7:59 PM

8081?

A

ashley•3/1/24, 8:01 PM

No, wrong, see above.
2024-03-01T16:08:54.761577365Z [FATAL tini (6)] exec docker failed: No such file or directory Error

A

ashley•3/1/24, 8:02 PM

Use EXTERNAL port NOT the internal one.

Aashley Use EXTERNAL port NOT the internal one.

V

voiceofkronosOP•3/1/24, 8:04 PM

okay but the endpoint is working on 8081, how am ı going to hit it? Should i bind it to 8081?

Vvoiceofkronos okay but the endpoint is working on 8081, how am ı going to hit it? Should i bin...

A

ashley•3/1/24, 8:07 PM

No, bind it to 0.0.0.0

A

ashley•3/1/24, 8:07 PM

Hit it by connecting to EXTERNAL PORT.

A

ashley•3/1/24, 8:08 PM

Use TCP port mapping, or else change it to an HTTP port and use the proxy URL

A

ashley•3/1/24, 8:08 PM

You CANNOT connect to 8081, its impossible. EXTERNAL PORT only. NOT INTERNAL.

V

voiceofkronosOP•3/1/24, 8:23 PM

solved, thanks for your support!! @ashleyk

A

ashley•3/1/24, 8:25 PM

Glad you got it working