MinIO

Overview

Export evaluation data to MinIO, or any other S3-compatible object store such as Ceph or Garage. There is no dedicated MinIO exporter: you use the S3 exporter and point it at your own server.

MinIO is an S3-compatible object store you can run yourself, on-premises or in your own cloud. Everything below uses the AWS S3 exporter with kind: s3 — only the endpoint changes.

Everytime the flushInterval or maxEventInMemory is reached, a new file is added to your MinIO bucket.

Configure the relay proxy

The exporter configuration is the regular s3 one, MinIO is selected through the AWS SDK environment variables:

goff-proxy.yaml

# ...
exporters:
  - kind: s3
    bucket: evaluation-data-bucket
    path: /go-feature-flag/variations/
# ...

export AWS_ENDPOINT_URL_S3=http://127.0.0.1:9000
export AWS_ACCESS_KEY_ID=minioadmin
export AWS_SECRET_ACCESS_KEY=minioadmin
export AWS_DEFAULT_REGION=us-east-1

Only kind and bucket are mandatory. All the other fields (format, filename, path, flushInterval, maxEventInMemory, csvTemplate, parquetCompressionCodec) behave exactly as documented on the AWS S3 exporter page — MinIO changes nothing about them.

Use an IP address in AWS_ENDPOINT_URL_S3, or read below

The AWS SDK addresses buckets as http://{bucket}.{your-endpoint} (virtual-host style), except when your endpoint is an IP address. So http://127.0.0.1:9000 works out of the box, but http://minio:9000 makes the relay proxy call http://evaluation-data-bucket.minio:9000 and fail.

The relay proxy has no setting to force path-style addressing today. If you cannot use an IP address, set MINIO_DOMAIN on your MinIO server and make {bucket}.{host} resolve to it — a full Docker Compose example is on the MinIO retriever page.

Configure the GO Module

The GO Module exposes S3ClientOptions, which lets you force path-style addressing. This works with any endpoint, hostname or IP, and requires no MinIO-side configuration:

example.go

awsConfig, _ := config.LoadDefaultConfig(context.Background(),
    config.WithBaseEndpoint("http://minio:9000"))

config := ffclient.Config{
    // ...
    DataExporter: ffclient.DataExporter{
        // ...
        Exporter: &s3exporterv2.Exporter{
            Format:    "json",
            Bucket:    "evaluation-data-bucket",
            S3Path:    "/go-feature-flag/variations/",
            Filename:  "flag-variation-{{ .Timestamp}}.{{ .Format}}",
            AwsConfig: &awsConfig,
            S3ClientOptions: []func(*s3.Options){
                func(o *s3.Options) { o.UsePathStyle = true },
            },
        },
    },
    // ...
}
err := ffclient.Init(config)
defer ffclient.Close()

Field	Mandatory	Description
Bucket		The name of your bucket.
AwsConfig		An instance of aws.Config. Set your MinIO endpoint with config.WithBaseEndpoint(). If you omit it, config.LoadDefaultConfig() is called for you, which reads AWS_ENDPOINT_URL_S3 and the other standard AWS environment variables.
S3ClientOptions		Functional options passed to the S3 client. Set o.UsePathStyle = true to address buckets as {endpoint}/{bucket}, which is what MinIO expects by default.

info

AwsConfig is optional. If you leave it out, GO Feature Flag calls config.LoadDefaultConfig() itself, so setting AWS_ENDPOINT_URL_S3, AWS_ACCESS_KEY_ID, AWS_SECRET_ACCESS_KEY and AWS_DEFAULT_REGION in your environment is enough — you only need S3ClientOptions to force path-style addressing.

The remaining fields are shared with the AWS S3 exporter.

Troubleshooting

NoSuchBucket: The specified bucket does not exist — but the bucket does exist

Your endpoint is a hostname, so the SDK is using virtual-host addressing and MinIO reads the first path segment as the bucket name. Use an IP endpoint, configure MINIO_DOMAIN and a DNS alias, or set UsePathStyle if you are on the GO Module.

resolve auth scheme: resolve endpoint: endpoint rule error, Invalid region

The AWS SDK requires a region even when talking to MinIO. Any value works:

export AWS_DEFAULT_REGION=us-east-1

WARN Response has no supported checksum. Not validating response payload.

Harmless, but if it floods your logs you can silence it with:

export AWS_RESPONSE_CHECKSUM_VALIDATION=when_required