# MinIO
[](https://hackmd.io/@Chieh) [](https://github.com/chiehpower/Setup-deeplearning-tools/tree/master/minIO) [](https://hub.docker.com/layers/minio/minio/minio/RELEASE.2022-03-24T00-43-44Z/images/sha256-ffbd26fc05e4d7423421486f0433f9d4bd27e3ea38b6c02d68d8d5bcac75af67?context=explore)

## Purpose

Integrate MinIO into Triton inference server

## Introduction

Dashboard Layout:
- Buckets
- Identity
- Users
- Groups
- Service Accounts
- Access
- Monitoring
- Metrics
- Logs
- Audit
- Trace
- Watch
- Drives
- Support
- Register
- Health
- Performance
- License
- Settings
- Configurations
- Notifications
- Tiers
- Documentation
- Logout

## Deployment

Create two folders for storing the data from the container.
- data
- config

1. Pull the image first:

    docker pull minio/minio:RELEASE.2022-02-01T18-00-14Z 

2. Launch a container:

    docker run -p 9000:9000 --name minio1 \
                -v data:/data \
                -v config:/root/.minio \
                minio/minio:RELEASE.2022-02-01T18-00-14Z \
                server /data --console-address ":9001"

Or you can use `docker-compose up` from my [docker-compose file](https://github.com/chiehpower/Setup-deeplearning-tools/tree/master/minIO)

Access the link by `0.0.0.0:9000` and then you can see the console.

- Password : minioadmin
- User name : minioadmin



## Basic MinIO usage

Please check my instruction [here](https://github.com/chiehpower/Setup-deeplearning-tools/tree/master/minIO).

### Reference

- [安裝MinIO並從notebook儲存model到MinIO](https://ithelp.ithome.com.tw/articles/10275077?sc=hot)
- [Github: ithome-ironman](https://github.com/masonwu1762/ithome-ironman)

---

## Start Triton Server with MinIO as model repository

### Notes:

1. I set the MINIO_ROOT_USER and MINIO_ROOT_PASSWORD in the docker compose file.

- MINIO_ROOT_USER: user
- MINIO_ROOT_PASSWORD: user123456
2. I set the default bucket that it will automatically create a folder after it starts the container.
3. Triton will connect to the model bucket of local minIO container.
4. No need to set up everything on AWS cloud services. Everything is on host.
5. You can download the testing model via the script of `fetch_models.sh` file.

### Usage

One command:`docker-compose -f docker-compose-triton.yml up`


### Reference

- [Deploy Nvidia Triton Inference Server with MinIO as Model Store](https://thenewstack.io/deploy-nvidia-triton-inference-server-with-minio-as-model-store/)