To run this example, you need Kubernetes version >=1.4 cluster installed and running, and that you have installed the [`kubectl`](https://kubernetes.io/docs/tasks/kubectl/install/) command line tool in your path. Please see the [getting started guides](https://kubernetes.io/docs/getting-started-guides/) for installation instructions for your platform.
## Minio Standalone Server Deployment
The following section describes the process to deploy standalone [Minio](https://minio.io/) server on Kubernetes. The deployment uses the [official Minio Docker image](https://hub.docker.com/r/minio/minio/~/dockerfile/) from Docker Hub.
This section uses following core components of Kubernetes:
Minio needs persistent storage to store objects. If there is no
persistent storage, the data stored in Minio instance will be stored in the container file system and will be wiped off as soon as the container restarts.
Create a persistent volume claim (PVC) to request storage for the Minio instance. Kubernetes looks out for PVs matching the PVC request in the cluster and binds it to the PVC automatically.
This is the PVC description.
```sh
apiVersion: v1
kind: PersistentVolumeClaim
metadata:
# This name uniquely identifies the PVC. This is used in deployment.
name: minio-pv-claim
spec:
# Read more about access modes here: http://kubernetes.io/docs/user-guide/persistent-volumes/#access-modes
accessModes:
# The volume is mounted as read-write by a single node
- ReadWriteOnce
resources:
# This is the request for storage. Should be available in the cluster.
A deployment encapsulates replica sets and pods—so, if a pod goes down, replication controller makes sure another pod comes up automatically. This way you won’t need to bother about pod failures and will have a stable Minio service available.
This is the deployment description.
```sh
apiVersion: extensions/v1beta1
kind: Deployment
metadata:
# This name uniquely identifies the Deployment
name: minio
spec:
strategy:
# Specifies the strategy used to replace old Pods by new ones
Now that you have a Minio deployment running, you may either want to access it internally (within the cluster) or expose it as a Service onto an external (outside of your cluster, maybe public internet) IP address, depending on your use case. You can achieve this using Services. There are 3 major service types—default type is ClusterIP, which exposes a service to connection from inside the cluster. NodePort and LoadBalancer are two types that expose services to external traffic.
In this example, we expose the Minio Deployment by creating a LoadBalancer service. This is the service description.
```sh
apiVersion: v1
kind: Service
metadata:
# This name uniquely identifies the service
name: minio-service
spec:
type: LoadBalancer
ports:
- port: 9000
targetPort: 9000
protocol: TCP
selector:
# Looks for labels `app:minio` in the namespace and applies the spec
You can update an existing Minio deployment to use a newer Minio release. To do this, use the `kubectl set image` command:
```sh
kubectl set image deployment/minio-deployment minio=<replace-with-new-minio-image>
```
Kubernetes will restart the deployment to update the image. You will get a message as shown below, on successful update:
```
deployment "minio-deployment" image updated
```
### Standalone Resource cleanup
You can cleanup the cluster using
```sh
kubectl delete deployment minio \
&& kubectl delete pvc minio-pv-claim \
&& kubectl delete svc minio-service
```
## Minio Distributed Server Deployment
The following document describes the process to deploy [distributed Minio](https://docs.minio.io/docs/distributed-minio-quickstart-guide) server on Kubernetes. This example uses the [official Minio Docker image](https://hub.docker.com/r/minio/minio/~/dockerfile/) from Docker Hub.
This example uses following core components of Kubernetes:
Headless Service controls the domain within which StatefulSets are created. The domain managed by this Service takes the form: `$(service name).$(namespace).svc.cluster.local` (where “cluster.local” is the cluster domain), and the pods in this domain take the form: `$(pod-name-{i}).$(service name).$(namespace).svc.cluster.local`. This is required to get a DNS resolvable URL for each of the pods created within the Statefulset.
A StatefulSet provides a deterministic name and a unique identity to each pod, making it easy to deploy stateful distributed applications. To launch distributed Minio you need to pass drive locations as parameters to the minio server command. Then, you’ll need to run the same command on all the participating pods. StatefulSets offer a perfect way to handle this requirement.
This is the Statefulset description.
```sh
apiVersion: apps/v1beta1
kind: StatefulSet
metadata:
# This name uniquely identifies the StatefulSet
name: minio
spec:
serviceName: minio
replicas: 4
selector:
matchLabels:
app: minio # has to match .spec.template.metadata.labels
template:
metadata:
labels:
app: minio # has to match .spec.selector.matchLabels
Now that you have a Minio statefulset running, you may either want to access it internally (within the cluster) or expose it as a Service onto an external (outside of your cluster, maybe public internet) IP address, depending on your use case. You can achieve this using Services. There are 3 major service types—default type is ClusterIP, which exposes a service to connection from inside the cluster. NodePort and LoadBalancer are two types that expose services to external traffic.
In this example, we expose the Minio Deployment by creating a LoadBalancer service. This is the service description.
On successful update, you should see the output below
```
statefulset "minio" patched
```
Then delete all the pods in your StatefulSet one by one as shown below. Kubernetes will restart those pods for you, using the new image.
```sh
kubectl delete minio-0
```
### Resource cleanup
You can cleanup the cluster using
```sh
kubectl delete statefulset minio \
&& kubectl delete svc minio \
&& kubectl delete svc minio-service
```
### Deploying on cluster nodes with local host path
If your cluster does not have a storage solution or PV abstraction, you must explicitly define what nodes you wish to run Minio on, and define a homogeneous path to a local fast block device available on every host.
This must be changed in the example daemonset: [minio-distributed-daemonset.yaml](minio-distributed-daemonset.yaml)
Specifically the hostpath:
```yaml
hostPath:
path: /data/minio/
```
And the list of hosts:
```yaml
- http://hostname1:9000/data/minio
- http://hostname2:9000/data/minio
- http://hostname3:9000/data/minio
- http://hostname4:9000/data/minio
```
Once deployed, tag the defined host with the `minio-server=true` label:
```bash
kubectl label node hostname1 -l minio-server=true
kubectl label node hostname2 -l minio-server=true
kubectl label node hostname3 -l minio-server=true
kubectl label node hostname4 -l minio-server=true
```
## Minio GCS Gateway Deployment
The following section describes the process to deploy [Minio](https://minio.io/) GCS Gateway on Kubernetes. The deployment uses the [official Minio Docker image](https://hub.docker.com/r/minio/minio/~/dockerfile/) from Docker Hub.
This section uses following core components of Kubernetes:
Create the Google Cloud Service credentials file using the steps mentioned [here](https://github.com/minio/minio/blob/master/docs/gateway/gcs.md#create-service-account-key-for-gcs-and-get-the-credentials-file).
Use the path of file generated above to create a Kubernetes `secret`.
A `secret` is intended to hold sensitive information, such as passwords, OAuth tokens, and ssh keys. Putting this information in a secret is safer and more flexible than putting it verbatim in a pod definition or in a docker image.
Create the Google Cloud Service credentials file using the steps mentioned [here](https://github.com/minio/minio/blob/master/docs/gateway/gcs.md#create-service-account-key-for-gcs-and-get-the-credentials-file).
Use the path of file generated above to create a Kubernetes `secret`.
A deployment encapsulates replica sets and pods—so, if a pod goes down, replication controller makes sure another pod comes up automatically. This way you won’t need to bother about pod failures and will have a stable Minio service available.
Minio Gateway uses GCS as its storage backend and need to use a GCP `projectid` to identify your credentials. Update the section `gcp_project_id` with your
GCS project ID. This is the deployment description.
Now that you have a Minio deployment running, you may either want to access it internally (within the cluster) or expose it as a Service onto an external (outside of your cluster, maybe public internet) IP address, depending on your use case. You can achieve this using Services. There are 3 major service types—default type is ClusterIP, which exposes a service to connection from inside the cluster. NodePort and LoadBalancer are two types that expose services to external traffic.
In this example, we expose the Minio Deployment by creating a LoadBalancer service. This is the service description.
Kubernetes concepts like Deployments and StatefulSets provide perfect platform to deploy Minio server in standalone, distributed or shared mode. There are multiple options to deploy Minio on Kubernetes, you can choose the one that suits you the most.
Kubernetes concepts like Deployments and StatefulSets provide perfect platform to deploy Minio server in standalone, distributed or gateway mode. There are multiple options to deploy Minio on Kubernetes, you can choose the one that best suits your requirements.
- Minio [Helm](https://helm.sh) Chart offers a customizable and easy Minio deployment, with a single command. Read more about Minio Helm deployment [here](#prerequisites).
- Helm Chart: Minio Helm Chart offers customizable and easy Minio deployment with a single command. Refer [Minio Helm Chart repository documentation](https://github.com/helm/charts/tree/master/stable/minio) for more details.
- You can also explore Kubernetes [Minio example](https://github.com/minio/minio/blob/master/docs/orchestration/kubernetes-yaml/README.md) to deploy Minio using `.yaml` files.
- YAML File: Minio can be deployed with `yaml` files via `kubectl`. This document outlines steps required to deploy Minio using `yaml` files.
- If you'd like to get started with Minio on Kubernetes without having to create a real container cluster, you can also [deploy Minio locally](https://raw.githubusercontent.com/minio/minio/master/docs/orchestration/minikube/README.md) with MiniKube.
## Table of Contents
<aname="prerequisites"></a>
## 1. Prerequisites
- [Prerequisites](#prerequisites)
- [Minio Standalone Server Deployment](#minio-standalone-server-deployment)
Above command deploys Minio on the Kubernetes cluster in the default configuration. Below section lists all the configurable parameters of the Minio chart and their default values.
You can specify each parameter using the `--set key=value[,key=value]` argument to `helm install`. For example,
To run this example, you need Kubernetes version >=1.4 cluster installed and running, and that you have installed the [`kubectl`](https://kubernetes.io/docs/tasks/kubectl/install/) command line tool in your path. Please see the [getting started guides](https://kubernetes.io/docs/getting-started-guides/) for installation instructions for your platform.
```bash
$ helm install --name my-release \
--set persistence.size=100Gi \
stable/minio
## Minio Standalone Server Deployment
The following section describes the process to deploy standalone [Minio](https://minio.io/) server on Kubernetes. The deployment uses the [official Minio Docker image](https://hub.docker.com/r/minio/minio/~/dockerfile/) from Docker Hub.
This section uses following core components of Kubernetes:
The above command deploys Minio server with a 100Gi backing persistent volume.
### Create Persistent Volume Claim
Minio needs persistent storage to store objects. If there is no
persistent storage, the data stored in Minio instance will be stored in the container file system and will be wiped off as soon as the container restarts.
Create a persistent volume claim (PVC) to request storage for the Minio instance. Kubernetes looks out for PVs matching the PVC request in the cluster and binds it to the PVC automatically.
This is the PVC description.
```sh
apiVersion: v1
kind: PersistentVolumeClaim
metadata:
# This name uniquely identifies the PVC. This is used in deployment.
name: minio-pv-claim
spec:
# Read more about access modes here: http://kubernetes.io/docs/user-guide/persistent-volumes/#access-modes
accessModes:
# The volume is mounted as read-write by a single node
- ReadWriteOnce
resources:
# This is the request for storage. Should be available in the cluster.
requests:
storage: 10Gi
```
Alternately, you can provide a YAML file that specifies parameter values while installing the chart. For example,
A deployment encapsulates replica sets and pods—so, if a pod goes down, replication controller makes sure another pod comes up automatically. This way you won’t need to bother about pod failures and will have a stable Minio service available.
This is the deployment description.
```sh
apiVersion: extensions/v1beta1
kind: Deployment
metadata:
# This name uniquely identifies the Deployment
name: minio
spec:
strategy:
# Specifies the strategy used to replace old Pods by new ones
# This label is used as a selector in Service definition
app: minio
spec:
# Volumes used by this deployment
volumes:
- name: data
# This volume is based on PVC
persistentVolumeClaim:
# Name of the PVC created earlier
claimName: minio-pv-claim
containers:
- name: minio
# Volume mounts for this container
volumeMounts:
# Volume 'data' is mounted to path '/data'
- name: data
mountPath: "/data"
# Pulls the lastest Minio image from Docker Hub
image: minio/minio:RELEASE.2018-10-18T00-28-58Z
args:
- server
- /data
env:
# Minio access key and secret key
- name: MINIO_ACCESS_KEY
value: "minio"
- name: MINIO_SECRET_KEY
value: "minio123"
ports:
- containerPort: 9000
# Readiness probe detects situations when Minio server instance
# is not ready to accept traffic. Kubernetes doesn't forward
# traffic to the pod till readiness checks fail.
readinessProbe:
httpGet:
path: /minio/health/ready
port: 9000
initialDelaySeconds: 120
periodSeconds: 20
# Liveness probe detects situations where Minio server instance
# is not working properly and needs restart. Kubernetes automatically
# restarts the pods if liveness checks fail.
livenessProbe:
httpGet:
path: /minio/health/live
port: 9000
initialDelaySeconds: 120
periodSeconds: 20
```
This chart provisions a Minio server in standalone mode, by default. To provision Minio server in [distributed mode](https://docs.minio.io/docs/distributed-minio-quickstart-guide), set the `mode` field to `distributed`,
This provisions Minio server in distributed mode with 4 nodes. To change the number of nodes in your distributed Minio server, set the `numberOfNodes` field,
### Create Minio Service
Now that you have a Minio deployment running, you may either want to access it internally (within the cluster) or expose it as a Service onto an external (outside of your cluster, maybe public internet) IP address, depending on your use case. You can achieve this using Services. There are 3 major service types—default type is ClusterIP, which exposes a service to connection from inside the cluster. NodePort and LoadBalancer are two types that expose services to external traffic.
In this example, we expose the Minio Deployment by creating a LoadBalancer service. This is the service description.
```sh
apiVersion: v1
kind: Service
metadata:
# This name uniquely identifies the service
name: minio-service
spec:
type: LoadBalancer
ports:
- port: 9000
targetPort: 9000
protocol: TCP
selector:
# Looks for labels `app:minio` in the namespace and applies the spec
#### StatefulSet [limitations](http://kubernetes.io/docs/concepts/abstractions/controllers/statefulsets/#limitations) applicable to distributed Minio
### Update existing Minio Deployment
* StatefulSets need persistent storage, so the `persistence.enabled` flag is ignored when `mode` is set to `distributed`.
* When uninstalling a distributed Minio release, you'll need to manually delete volumes associated with the StatefulSet.
You can update an existing Minio deployment to use a newer Minio release. To do this, use the `kubectl set image` command:
### Shared Minio
```sh
kubectl set image deployment/minio-deployment minio=<replace-with-new-minio-image>
```
To provision Minio servers in [shared mode](https://github.com/minio/minio/blob/master/docs/shared-backend/README.md), set the `mode` field to `shared`,
Kubernetes will restart the deployment to update the image. You will get a message as shown below, on successful update:
```bash
$ helm install --set mode=shared stable/minio
```
deployment "minio-deployment" image updated
```
This provisions 4 Minio server nodes backed by single storage. To change the number of nodes in your shared Minio deployment, set the `numberOfNodes` field,
This provisions Minio server in shared mode with 8 nodes.
## Minio Distributed Server Deployment
### Persistence
The following document describes the process to deploy [distributed Minio](https://docs.minio.io/docs/distributed-minio-quickstart-guide) server on Kubernetes. This example uses the [official Minio Docker image](https://hub.docker.com/r/minio/minio/~/dockerfile/) from Docker Hub.
This chart provisions a PersistentVolumeClaim and mounts corresponding persistent volume to default location `/export`. You'll need physical storage available in the Kubernetes cluster for this to work. If you'd rather use `emptyDir`, disable PersistentVolumeClaim by:
This example uses following core components of Kubernetes:
Headless Service controls the domain within which StatefulSets are created. The domain managed by this Service takes the form: `$(service name).$(namespace).svc.cluster.local` (where “cluster.local” is the cluster domain), and the pods in this domain take the form: `$(pod-name-{i}).$(service name).$(namespace).svc.cluster.local`. This is required to get a DNS resolvable URL for each of the pods created within the Statefulset.
This is the Headless service description.
```sh
apiVersion: v1
kind: Service
metadata:
name: minio
labels:
app: minio
spec:
clusterIP: None
ports:
- port: 9000
name: minio
selector:
app: minio
```
> *"An emptyDir volume is first created when a Pod is assigned to a Node, and exists as long as that Pod is running on that node. When a Pod is removed from a node for any reason, the data in the emptyDir is deleted forever."*
You can update an existing Minio Helm Release to use a newer Minio Docker image. To do this, use the `helm upgrade` command:
### Create Minio Statefulset
A StatefulSet provides a deterministic name and a unique identity to each pod, making it easy to deploy stateful distributed applications. To launch distributed Minio you need to pass drive locations as parameters to the minio server command. Then, you’ll need to run the same command on all the participating pods. StatefulSets offer a perfect way to handle this requirement.
This is the Statefulset description.
```sh
apiVersion: apps/v1beta1
kind: StatefulSet
metadata:
# This name uniquely identifies the StatefulSet
name: minio
spec:
serviceName: minio
replicas: 4
selector:
matchLabels:
app: minio # has to match .spec.template.metadata.labels
template:
metadata:
labels:
app: minio # has to match .spec.selector.matchLabels
Now that you have a Minio statefulset running, you may either want to access it internally (within the cluster) or expose it as a Service onto an external (outside of your cluster, maybe public internet) IP address, depending on your use case. You can achieve this using Services. There are 3 major service types—default type is ClusterIP, which exposes a service to connection from inside the cluster. NodePort and LoadBalancer are two types that expose services to external traffic.
In this example, we expose the Minio Deployment by creating a LoadBalancer service. This is the service description.
On successful update, you should see the output below
```bash
Release "your-helm-release" has been upgraded. Happy Helming!
```
statefulset "minio" patched
```
## 4. Uninstalling the Chart
Then delete all the pods in your StatefulSet one by one as shown below. Kubernetes will restart those pods for you, using the new image.
Assuming your release is named as `my-release`, delete it using the command:
```sh
kubectl delete minio-0
```
```bash
$ helm delete my-release
### Resource cleanup
You can cleanup the cluster using
```sh
kubectl delete statefulset minio \
&& kubectl delete svc minio \
&& kubectl delete svc minio-service
```
The command removes all the Kubernetes components associated with the chart and deletes the release.
### Deploying on cluster nodes with local host path
### Notes
If your cluster does not have a storage solution or PV abstraction, you must explicitly define what nodes you wish to run Minio on, and define a homogeneous path to a local fast block device available on every host.
* An instance of a chart running in a Kubernetes cluster is called a release. Helm automatically assigns a unique release name after installing the chart. You can also set your preferred name by:
This must be changed in the example daemonset: [minio-distributed-daemonset.yaml](minio-distributed-daemonset.yaml)
```bash
$ helm install --name my-release stable/minio
Specifically the hostpath:
```yaml
hostPath:
path: /data/minio/
```
* To override the default keys, pass the access and secret keys as arguments to helm install.
And the list of hosts:
```yaml
- http://hostname1:9000/data/minio
- http://hostname2:9000/data/minio
- http://hostname3:9000/data/minio
- http://hostname4:9000/data/minio
```
Once deployed, tag the defined host with the `minio-server=true` label:
The following section describes the process to deploy [Minio](https://minio.io/) GCS Gateway on Kubernetes. The deployment uses the [official Minio Docker image](https://hub.docker.com/r/minio/minio/~/dockerfile/) from Docker Hub.
This section uses following core components of Kubernetes:
Create the Google Cloud Service credentials file using the steps mentioned [here](https://github.com/minio/minio/blob/master/docs/gateway/gcs.md#create-service-account-key-for-gcs-and-get-the-credentials-file).
Use the path of file generated above to create a Kubernetes `secret`.
A `secret` is intended to hold sensitive information, such as passwords, OAuth tokens, and ssh keys. Putting this information in a secret is safer and more flexible than putting it verbatim in a pod definition or in a docker image.
Create the Google Cloud Service credentials file using the steps mentioned [here](https://github.com/minio/minio/blob/master/docs/gateway/gcs.md#create-service-account-key-for-gcs-and-get-the-credentials-file).
Use the path of file generated above to create a Kubernetes `secret`.
A deployment encapsulates replica sets and pods—so, if a pod goes down, replication controller makes sure another pod comes up automatically. This way you won’t need to bother about pod failures and will have a stable Minio service available.
Minio Gateway uses GCS as its storage backend and need to use a GCP `projectid` to identify your credentials. Update the section `gcp_project_id` with your
GCS project ID. This is the deployment description.
Now that you have a Minio deployment running, you may either want to access it internally (within the cluster) or expose it as a Service onto an external (outside of your cluster, maybe public internet) IP address, depending on your use case. You can achieve this using Services. There are 3 major service types—default type is ClusterIP, which exposes a service to connection from inside the cluster. NodePort and LoadBalancer are two types that expose services to external traffic.
In this example, we expose the Minio Deployment by creating a LoadBalancer service. This is the service description.
* Execute the `minio_distributed.sh` script in command prompt.
```sh
./minio_distributed.sh
```
After the script is executed successfully, you should get an output like this
```sh
service "minio-public" created
service "minio" created
statefulset "minio" created
```
This means Minio is deployed on your local Minikube installation.
Note that the service `minio-public` is a [clusterIP](https://kubernetes.io/docs/user-guide/services/#publishing-services---service-types) service. It exposes the service on a cluster-internal IP. To connect to your Minio instances via `kubectl port-forward` command, execute
```
kubectl port-forward minio-0 9000:9000
```
Minio server can now be accessed at `http://localhost:9000`, with accessKey and secretKey as mentioned in the `statefulset.yaml` file.
## 3. Notes
Minikube currently does not support dynamic provisioning, so we manually create PersistentVolumes(PV) and PersistentVolumeClaims(PVC). Once the PVs and PVCs are created, we call the `statefulset.yaml` configuration file to create the distributed Minio setup.
This setup runs on a laptop/computer. Hence only one disk is used as the backend for all the minio instance PVs. Minio sees these PVs as separate disks and reports the available storage incorrectly.
Nitish Tiwari
6 years ago
committed by