Hi team, I'm having a persistent issue trying to deploy LakeFS to an EKS cluster using the Terraform

helm_release

resource, and I'm hoping someone might have some insight. Passing configuration via `values`: I've tried passing the configuration using both the

lakefsConfig

key and the

config

key (as shown on Artifact Hub). In both cases,

helm get values lakefs

confirms that Helm receives the correct values from Terraform. However, the resulting

ConfigMap

in the cluster is still the default one.

Hi @Jose Ignacio Gascon Conde and welcome to the lake lakefs Are you working with our documentation?

Just wanted to share a quick summary of a challenging (but ultimately successful!) deployment of LakeFS on EKS, in case my experience helps anyone else in the community. My setup: I'm managing my infrastructure with Terraform, so I was using the

helm_release

resource to deploy the LakeFS image

v1.29.0

with my own RDS and S3. The mystery I hit: I was scratching my head for a while because my

config.yaml

inside the pod was always stuck on the default

type: local

settings. The weird part was that

helm get values

showed that my configuration was being passed to Helm perfectly. The rabbit hole: I went down a deep rabbit hole and found a tricky catch-22. • I figured out I needed to use the

lakefsConfig

key for the pod to pick up the config file. • But, as soon as I added

lakefsConfig

, the deployment would consistently fail with a

fromYaml

type error coming from the

_s3proxyContainer.tpl

template (I tested this with chart

v1.3.3

and newer). So I was stuck: I couldn't configure the app without

lakefsConfig

, but I couldn't deploy if I used it. Just to be clear, this all seems to stem from the interaction between the chart and the Terraform

helm_release

provider. It’s very possible that a direct

helm install -f values.yaml

would have worked fine, but my entire workflow is based on Terraform. What finally worked for me: I decided to stop fighting the chart's templating logic. My solution was to: 1. Abandon the

helm_release

resource entirely. 2. Use

helm template

to render the raw Kubernetes YAMLs locally. 3. Translate those manifests into native Terraform HCL resources (

kubernetes_deployment

,

kubernetes_config_map

, etc.). This gave me direct control to build and mount the

ConfigMap

exactly as needed and bypassed all the issues. Anyway, just wanted to put this out there as feedback and in case it saves someone else a few hours of debugging. The tool is awesome, and I'm happy to be up and running now