Feature server has aws lambda and gcp cloudrun deployment options, lambda being marked alpha and cloudrun undocumented and probably not functional. Both are unmaintained at this point. Their deployment is opinionated and managed by feast cli itself, in other words they are deployed with a feast apply command, which I think makes them very impractical for a real-world infra deployment. It's practically impossible to deploy them as part of some sort of IaC workflow.
On the flip side, we have to maintain a whole bunch of code that does a lot of vendor-specific infra management, for example lambda deployment contains uploading docker image do ecr, communicating with aws services using boto client and so on (aws.py). There is parallel work going on supporting an official feast helm chart, which looks like a more stable alternative to me. I propose to remove lambda and cloudrun deployment options from feast.
Feature server has aws lambda and gcp cloudrun deployment options, lambda being marked alpha and cloudrun undocumented and probably not functional. Both are unmaintained at this point. Their deployment is opinionated and managed by
feastcli itself, in other words they are deployed with afeast applycommand, which I think makes them very impractical for a real-world infra deployment. It's practically impossible to deploy them as part of some sort of IaC workflow.On the flip side, we have to maintain a whole bunch of code that does a lot of vendor-specific infra management, for example lambda deployment contains uploading docker image do ecr, communicating with aws services using boto client and so on (aws.py). There is parallel work going on supporting an official feast helm chart, which looks like a more stable alternative to me. I propose to remove lambda and cloudrun deployment options from feast.