Runpod

We're a community of enthusiasts, engineers, and enterprises, all sharing insights on AI, Machine Learning and GPUs!

Join

Runpod

We're a community of enthusiasts, engineers, and enterprises, all sharing insights on AI, Machine Learning and GPUs!

Join

⚡｜serverless

⛅｜pods

🔧｜api-opensource

📡｜instant-clusters

🗂｜hub

nimbot

9/18/2025

Requests not leaving queue

Hello! I have a serverless endpoint mm45qixfgrgydr.. All tasks remain in queue, while workers are running and without working on tasks in the queue. Can you help me resolve this? Obviously all the requests eventually timeout

PavloPalyz

9/17/2025

Request in queue are not processed

Sometimes the serverless endpoint doesn't start to process requests in queue, until the request count to start a second worker is reached. If this is a feature, how do I turn it off? And if it's a bug, how do I resolve it?...

CamaroZ

9/17/2025

Unable use model gemma3:27b-it-q4_K_M

Hello guys, i`m trying deploy a serverless function with model gemma3:27b-it-q4_K_M and ollama 0.11.10, but when download is completed to my network storage isnt save to mounted location. Anyone have a issue like that? This screenshot shows the dir of blob and manifest....

Magnus

9/17/2025

Using secrets in serverless

Trying to create a vLLM serverless endpoint, but I want to use the HF_TOKEN from my secrets. How can I do that?

Backai

9/17/2025

High serverless costs despite no traffic

Hey, has anyone experienced this? One of our serverless endpoints has skyrocketed in costs within the last 5 days. However, there were virtually no requests running on that endpoint during this time period. How can this be? There was a surge in requests before the timeframe, but it seems those requests were billed correctly....

Accai

9/17/2025

Questions about S3 Image Uploading

Hey! I saw in the documentation that with serverless endpoints, you can send files directly to S3. It works well, but I get the impression that you can't specify the file name, path, or metadata. Is this possible and I just haven't found it, or do I have to manage this myself in the image?...

Accai

9/17/2025

Linking the upload of the image generated by comfy to Cloudflare R2 via S3 does not work

Hello, I am trying to upload images generated by my comfyui-based serverless endpoint to my Cloudflare R2 via s3Config. I did not put s3 config in the body of the request, I simply added the environment variables as indicated in the documentation. My bucket is in the European Union, and I created a special administrator API access for RunPod. I entered everything in the env variables. Despite this, I get a “Bucket not found” error and I can't fix the problem. Also, is there a way to test /run /runsync etc. locally to avoid burning through my credits?...

Solution:

Reason found ✅ For Cloudflare R2, you must specify the bucket name in the endpoint URL in addition to the bucket_name environment variable

billchen

9/17/2025

Can serverless applications be rolled back to a previous version?

Can serverless applications be rolled back to a previous version? The previous version was working fine, but after I updated the code and repackaged the Docker image, issues arose. I want to roll back to the previous version.

WeamonZ

9/16/2025

Docker & ComfyUI

Thread tutorial to understand the power of Docker & add comfyui to it. Might include how to deploy to runpod (pods & serverless) later....

Toseidon

9/16/2025

timeout

Hello! I have an serverless endpoint 538lzj0v01eu90 which previously worked fine. Tasks piled up in the queue and not being worked on despite there being 1-3 running workers. All tasks remain in queue, while workers are running and without working on tasks in the queue. This ofc results in timeout issues, while queue is still stuck. Can you help me resolve this?

Solution:

I don't see that this endpoint received traffic very often enough for me to help establish a baseline of if it worked before - however:

requirement error: unsatisfied condition: cuda>=12.8, please update your driver to a newer version, or use an earlier cuda container: unknown

requirement error: unsatisfied condition: cuda>=12.8, please update your driver to a newer version, or use an earlier cuda container: unknown

...

Accai

9/16/2025

Custom checkpoint on Comfy UI Serverless

What is the best way to use a custom checkpoint/loras/custom nodes on a serverless comfyUI?

Accai

9/16/2025

Worker in Idle despite configuration

Hey! I'm new here. What does it mean when a worker is “Idle,” even though I have configured everything so that it is idle as little as possible? Do I pay all the time or only when it is idle?

Solution:

👋 hey hey! Someone else might be able to add here but I can add a few things Yeah the idle workers thing might be a little confusing In that UI you should see a couple of different states, if workers are listed as idle then you are not being billed. Idle in this context just means they are initialized with your image and ready to serve requests. A "running" state is what you are billed for....

Five

9/16/2025

How to do auto scaling based on request spikes?

I am using 10 max workers for now, but sometimes i get spike in request. Which lead the request to stay in runpod queue. How to get rid of that?

chengxi

9/16/2025

Why is building a serverless endpoint so slow?

Why is building a serverless endpoint so slow? I wrote a Dockerfile based on Docker's worker-comfyui. It only installs a custom node and downloads the model. Each build takes half an hour, with a large amount of time spent in "#18 sha256: ... 8.51GB / 14.93GB 168.5s"

M-S

9/15/2025

Issues Accessing FastAPI /docs and /health via Proxy

I am running a FastAPI application inside my RunPod GPU pod. The application works correctly inside the container — for example: curl http://127.0.0.1:8000/upscale ...

Nikita

9/15/2025

Is anyone else facing serious outages today?

For the past 12-16 hours, we have been facing outages and throttling for about 10-15% of our serverless workers (mostly on the EU-SE-1 cluster).

Pronob saha

9/13/2025

Delay and Workers Not Starting on RunPod Serverless (SGLang Deployment,medgemma 4b VLM model)

I attempted to deploy my model using RunPod Serverless with SGLang, but I am facing an issue. The deployment starts, however, it only shows continuous delay and no workers are being started. As a result, requests are not being executed. Could you please check what might be causing this issue and suggest how I can resolve it?...

toantran.mei (Albert)

9/12/2025

error creating container: nvidia-smi: parsing output of line 0: failed to parse (pcie.link.gen.max)

Hello Runpod Team, I stuck error workder running but failed. I lost money but can't use service.. ```...

WeamonZ

9/12/2025

Serverless pod error 500 when uploading build

Hey, I get the following error during the upload stage of my runpod serverless build. Finishing 0--1 failed with 500, retrying Endpoint id : ljnx57qtdbr2oa Build id : f98bb0b5-7187-4c77-8207-606db3c556e8...

yolo

9/12/2025

Coldstart: docker private repo vs github

Hello, i had a problem with the cold start of a Docker container that I hosted from my private GitLab container registry and I was able to reduce the cold start time significantly when I moved the Dockerfile to GitHub and linked the repository to Runpod. I guess that the Docker image is built and stored at runpod. This means that cold starts are no longer a problem. Can I also upload finished containers directly to runpod, or are there other ways to optimize cold starts?...

Previous Next

Gaming

Programming

Runpod

We're a community of enthusiasts, engineers, and enterprises, all sharing insights on AI, Machine Learning and GPUs!

Runpod

We're a community of enthusiasts, engineers, and enterprises, all sharing insights on AI, Machine Learning and GPUs!