Up until now the Flyte backend you’ve been working with likely has been accessible only on
localhost and likely ran
entirely in one Docker container. In order to handle production load, and make use of all the additional features
Flyte offers, we’ll need to replace, add, and configure some components. This page describes at a high-level what a
production ready deployment might look like.
Usage of Helm¶
Flyte uses Helm to manage its deployment releases onto a K8s cluster. The chart and templates are located under the helm folder. There is a base
values.yaml file but there are several files that fine tune those settings.
values-eks.yamlshould be additionally applied for AWS EKS deployments.
values-gcp.yamlshould be additionally applied for GCP GKE deployments.
values-sandbox.yamlshould be additionally applied for our sandbox install. See the Sandbox Overview page for more information.
DataCatalog components rely on PostgreSQL to store persistent records. In the sandbox deployment, a containerized version of Postgres is included but for a proper Flyte installation, we recommend one of the cloud provided databases. For AWS, we recommend their RDS service, for GCP, Cloud SQL, and Azure, PostgreSQL.
Production Grade Object Store¶
Core Flyte components such as Admin, Propeller, and DataCatalog, as well as user runtime containers rely on an Object Store to hold files. The sandbox deployment comes with a containerized Minio, which offers AWS S3 compatibility. We recommend swapping this out for AWS S3 or GCP GCS.
As your Flyte user-base evolves, adding new projects is as simple as registering them through the command line
$ flytectl create project --id myflyteproject --name "My Flyte Project" --description "My very first project onboarding onto Flyte"
A cron which runs at the cadence specified in Flyte Admin configuration ensures that all Kubernetes resources necessary for the new project are created and new workflows can successfully be registered and executed within it. See flytectl for more information.
This project should immediately show up in the Flyte console after refreshing.
Cloud Based Pub/Sub Integration¶
Flyte relies on cloud-provided pub/sub and schedulers to provide automated periodic execution of your launch plans. In AWS,
Flyte ships with its own authorization server, as well as the ability to use an external authorization server if your external IDP supports it. See the authorization page for detailed configuration.