How to speed up creation of local deployment using atlas CLI for vector search

Jake_Turner · 2023-12-11T20:12:02.362Z

Hello, am trying to get a test environment going in GitHub actions for an application leveraging vector search.

As vector search is currently only supported in atlas you have to use Atlas CLI (docker image version doesnt seem to support creating local clusters). There is a GH actions that you can use for Atlas CLI but it currently only supports ubuntu. Atlas CLI local deploy currently only support MacOS and CentOS so that doesnt work.

You can just install with brew in a mac os based GH runner but I am getting a time to spin up the cluster of about 20 mins. Does anyone know any good ways to reduce this?

I am building a local cluster with default settings:
atlas deployments setup tester --type local --force

Thanks

Jakub_Lazinski · 2024-01-10T08:52:49.410Z

Hi @Jake_Turner ,

Thanks for raising this question.
One of our next priorities is to make the local Atlas experience work with GH Actions as well as take steps to improve the performance. In meantime, have you tried pulling the Atlas CLI docker image from your GH Action and run the local deployment the way you’d do normally with Docker as described on the other forum and in the docs here?

david_brewer · 2024-05-17T04:57:32.380Z

Hello,
to optimize the setup time for spinning up a local Atlas cluster using git hub actions on a macOS runner consider caching dependencies to avoid redundant installations utilizing parallel jobs for concurrent execution of tasks, and optimizing Docker images to minimize size and improve performance. Additionally creating a pre-built local cluster image and fine-tuning dependencies can contribute to faster deployment. Continuously monitor workflow performance to identify and address bottlenecks for further optimization.