Edit endpoint with new docker image
Is it possible to update the deployed endpoint with new docker image linked to template?
15 Replies
If you update the Template with a new image RunPod will reload with new workers, using the updated template. It is important to use a new version. i.e. image:v2 instead of image (which is really image:latest) otherwise RunPod cannot tell you want to use a new image.
ok thank you
Sorry, I'm not following. When I created the endpoint I specified a specific tagname from my private DockerHub. Then, I pushed a new tag but I don't see a way to tell RunPod to pull this new tag. Unless I create a new endpoint, which is very inconvenient because that changes the URL and then I also need to update my application.
What's the recommended approach to update the Endpoint Docker image?
Nevermind, I've just found out the "New Release" option in the endpoint.
enter your specifiic endpoint, then select in top right menu, there is a button to deploy new tag release
"new release"
or just edit your template that is used by endpoint
In the endpoint you did not select an image. You selected a template. See highlighted section in attached image.
Inside of the template, you specified an Container image. See highlighted section in attached image below.
If you edit your template and update the Container image, such as changing from to in my example. If you use an image name like or (both will use :latest) it will work when you 1st deploy it but it will not auto update after that, evenif :latest is updated online. The only way a Template can detect that a new version is available is for you to add a new unique Container image tag name to the Template.
If you are using someone else's template, i.e. Through Quick Deploy in Serverless you cannot modify the image for this. It is not your template to modify.
I understand, but that doesn't seem to be working realiably in my case.
I'm using a custom image on my private Docker Hub. I simply edit the container image by specifying a new tag (new version). But it doesn't work unless I terminate all workers manually or edit the endpoint to have max workers = 0 and then back again to the desired amount.
So, you are editing the template with a new tag (version) and you still have to set max workers = 0 to get them to reload? I've not seen that behavior. Have you reached out to RunPod support under Help/Contact on their site?
Yes, sometimes it works, but sometimes the workers just hang updating and I have to kill them. I've just started today with the platform and I'm doing some tests to see if it's worth to migrate from fly.io. Everything sounds great when reading the product pages and documentation, but I'm just wondering if it's a stable and mature service to migrate our production GPU workloads...
I think what you are experiencing is the exception rather than the norm. I've never had issues with deploying a new version. I personally have not deployed productions GPU workloads, currently trying to build something. The main reason I chose RunPod is their ability to scale from 0. Though there are many users who are deploying production workloads I cannot directly speak for their experience. Are you planning on doing a lot of upgrades in production?
We’re doing a rolling update, so if you have multiple workers, they’ll be gradually updated to prevent any service interruptions.
hi everyone, I tried to use this github repo https://github.com/runpod-workers/worker-a1111 and I did docker build . on my ubuntu vbox but it always gives errors. What is link for runpod auto1111 serverless docker image which I can edit? thanks
GitHub
GitHub - runpod-workers/worker-a1111: Automatic1111 serverless worker.
Automatic1111 serverless worker. . Contribute to runpod-workers/worker-a1111 development by creating an account on GitHub.
Is changing the image version in Edit Template the same as using the New Release dialog?
Yes