Demo Trident

Guide to show some features and advantages to use ONTAP as a backend storage for Kubernetes and OpenShift using the last NetApp Lab on Demand Using Trident with Kubernetes and ONTAP. This repository includes some scripts and Ansible playbooks to customize the lab. The demo is shown in 6 sections:

• Module 1: Deployment & Provisioning
• Module 2: Kubernetes Storage Provisioning with Trident 101
• Module 3: New tier application architecture with Kubernetes
• Module 4: Advanced NetApp Trident features
• Module 5: Configuration Management
• Summary

Making the demo

Open the putty console and connect to the kubernetes master node root@rhel3. Run the below commands:

git clone https://github.com/pablogarciaarevalo/demo-trident
cd demo-trident/making
chmod 744 *.sh
. ./configure_demo.sh
cd ../demo/

Open the slides in the browser http://rhel6.demo.netapp.com/demo.pdf.

Module 1: Deployment & Provisioning

Show the demo kubernetes cluster

• Objetive: Show the kubernetes cluster in order to understand the demo.

Go to slide 2

Run the below command:

kubectl get nodes -o wide

Ansible phases days 0 & 1: Deployment & Provisioning

• Objetive: An Infrastructure as a Code software like Ansible is needed to automate when using Kubernetes. Ansible Phases 0 & 1 automates device bring up.

Go to slide 3

Run the below commands and explain the Ansible playbook to automate the ONTAP provisioning:

./01_iaac_ansible_day0-1.sh

Module 2: Kubernetes Storage Provisioning with Trident 101

Install NetApp Trident

• Objetive: Show the steps to deploy NetApp Trident.

Go to slide 4

Run the below command:

./02_install_trident.sh

Create NetApp Trident backend

• Objetive: A Trident backend defines the relationship between Trident and a storage system. It tells Trident how to communicate with that storage system. Tasks to do by the IT Storage Administrator.

Run the below command:

./03_create_k8s_backends.sh

Show the Kubernetes Storage Classes

• Objetive: A StorageClass provides a way for administrators to describe the 'classes' of storage they offer. Different classes might map to quality-of-service levels, or to backup policies, or to arbitrary policies determined by the cluster administrators. Trident moves the ONTAP's advantages to the containers and kubernetes solutions.

Run the below commands:

./04_create_k8s_storageclasses.sh

Dynamic Kubernetes Persistent Volumes Provisioning with Trident

• Objetive: Persistent Volume Claims (or PVCs) are objects that request storage resources from a cluster. It shows how to run Read Write Many Persistent Volumes and Read Write Once Persistent Volumes.

PV Access Modes	Abbreviated	Volume	As known as
ReadWriteMany	RWX	NFS	NAS
ReadWriteOnce	RWO	iSCSI or NFS	SAN
ReadOnlyMany	ROX	iSCSI or NFS

Go to slide 5

Run the below command:

./05_create_pvc.sh

Create Pods and bind the previous Persistent Volumes

• Objetive: Show how to bind RWM and RWO Persistent Volumes to a PODs. Show how the volumes are mounting.

Go to slide 6

Run the below command, noting that the YAML files are scheduling manually the Pods in the node rhel1 instead of using the kubernetes scheduler.

./06_create_pods.sh

kubectl exec -it pvpod-nas1 mount | grep /data
kubectl exec -it pvpod-san1 mount | grep /data

Scale the Pods manually

• Objetive: Run similar pods binding the same PV on the same kubernetes worker and different workers.

Go to slide 7

Run the below commands:

kubectl get pods -o wide

./07_scale_pods_manually.sh

kubectl get pods -o wide

All the pods with RWM PV will be running regardless of the worker on which they are scheduled. Only the pods with RWO scheduled on the worker rhel1 will be running. There is a Multi-Attach error for the RWO volume mount in the pods which are not scheduled in the worker rhel1. The reason is because ReadWriteMany (RMX) access is for workers not for pods. Focus on that usually the applications need RWM or RWO access. It depends on the application but K8S needs unified storage.

Go to slide 8

Sidecar Pod

• Objetive: Run a pod with two containers, the first one writes data in a shared volume and the second one reads the data. ReadWriteMany access mode should fit better.

Go to slide 9

Run the below command:

./08_sidecar_pod.sh

PVC RWO using NAS

• Objetive: Read Write Only access mode can be block (iSCSI) or file (NFS). Some applications don't run fast in NFS, like Kafka, but it's not an issue in the NAS side. A NAS is not slower.

Go to slide 10

Run the below command:

./09_pod_pvc_rwo.sh

Module 3: New tier application architecture with Kubernetes

Sample cloud-native application with 10 microservices by GCP

• Objetive: Data storage is not really important, the applications are. But Trident with the NetApp storage provides some advantages with the ONTAP features that allows the application works better (efficiency, scale, security, portability,...).

Go to slide 11

Set focus on the microservices Frontend and Cache (Redis), which can use RWM PV and RWO PV respectively. Again, K8S needs unified storage.

Kubernetes controllers: Deployment vs StatefulSet

• Objetive: Explain the differences between Deployment and StatefulSet. Like a Deployment, a StatefulSet manages Pods that are based on an identical container spec. Unlike a Deployment, a StatefulSet maintains a sticky identity for each of their Pods. These pods are created from the same spec, but are not interchangeable: each has a persistent identifier that it maintains across any rescheduling.

Go to slide 12

Keys:

• Stable, unique network identifiers
• Stable, persistent storage
• Ordered, graceful deployment and scaling
• Ordered, automated rolling updates

Create Kubernetes dummy frontend statefulset with ReadWriteMany Persistent Volume

• Objetive: Show how to create and scale a frontend deployment accessing a single ReadWriteMany Persistent Volume (NFS)

Go to slide 13 and 14

Run the below commands:

./10_create_frontend_service.sh

kubectl get pods -o wide
kubectl get pvc
kubectl get pv

Scaling the statefulset:

kubectl scale --replicas=5 deployment frontend
kubectl get pods -o wide

Create Kubernetes dummy backend statefulset with ReadWriteOnce Persistent Volume

• Objetive: Show how to create and scale a backend deployment accessing a single ReadWriteOnce Persistent Volume (iSCSI)

Go to slide 15

Run the below commands:

./11_create_backend_service.sh

kubectl get pods
kubectl get pvc

Scaling the statefulset:

kubectl scale --replicas=5 statefulset mongodb
kubectl get pods 

Module 4: Advanced NetApp Trident features

Volume Import

• Objetive: Explain the Trident Volume Import feature and some use cases. The demo shows a legacy website with the content in a NetApp NFS volume, and the procedure to import the webpage in Kubernetes as a pod. Note that the demo shows a website to understand it better but consider a database.

Go to slide 16 and 17

Open a browser http://rhel6.demo.netapp.com/

Run the below commands:

./12_import_web_service.sh

kubectl get pvc -n web

Run the below command to get the kubernetes services IP address:

kubectl get all -n web

Open a browser http://192.168.0.142

Go to slide 18

Scale the statefulset:

kubectl scale --replicas=3 statefulset web-prod -n web

Decoupling application and data with external storage

• Objetive: Kubernetes stateful applications are based on yaml files and persistent volumes. Using an external data storage allows to decoupling application and data. In case a failure, the service can be restored safely and quickly importing the persistent volumes and applying the yaml files. The most common cause of data and service lost is the human error. For instance, kubernetes doesn't request a --force flag or similar to delete a non-empty namespace (kubernetes/kubectl#525).

Run the below commands:

kubectl delete namespace web
kubectl get all -n web
kubectl get pvc --all-namespaces

Open a browser http://192.168.0.142

Having an external storage allows us to restore the service easily. Run the below commands:

./13_restore_web_service.sh
kubectl get all -n web

Open a browser http://192.168.0.142

Kubernetes Volume Snapshots

• Objetive: The container storage interface (CSI) is a standardized API for container orchestrators to manage storage plugins. NetApp Trident has been deployed a CSI plugin. Kubernetes 1.12 includes Volume Snapshots as a Alpha, and Kubernetes 1.17 does it as a Beta. This is CSI, but if you are using ONTAP you can have up to 1023 Snapshots per volume.

Go to slide 19

Kubernetes Persisten Volume Claim from Snapshot (aka Clone)

• Objetive: Note that right now, up to kubernetes 1.17, a Volume snapshot can not be restored. What can we do with a volume snapshot? Cloning.

Go to slide 20

Use case 1 for clone

• Objetive: Data pipeline for Kubeflow. Data Engineer preparing the data with SMB/NFS

Go to slide 21

./14_datalake_clone.sh

kubectl get pvc -n data-analytics

Go to slide 22

./15_data_analytics_services.sh

kubectl get all -n data-analytics

Use case 2 for clone

• Objetive: Coming to the previous web example. Cloning for dev/staging using Rest API

Go to slide 23

./16_clone_web.sh

Go to slide 24

GO TO SECOND K8S CLUSTER (root@rhel4):

cd demo-trident/demo/
./17_create_staging_web_service.sh
kubectl get all -n web

Open a browser http://192.168.0.150

./18_coding_new_website.sh

Open a browser using incognito mode http://192.168.0.150

Module 5: Configuration Management

Ansible phase 2: Configuration Management

• Objetive: Ansible Phases 2 optimizes compliance and operation.

Go to slide 25

GO BACK TO THE FIRST K8S CLUSTER (rhel3) and run the below command:

./19_iaac_ansible_day2.sh

Run the below command to show the volume protection with SnapMirror:

./20_get_protection_volumes.sh

Summary

• Objetive: Explain all the tasks done and resume the advantages of use NetApp storage

Go to slide 26

Open a browser http://192.168.0.141 (user: admin / password admin)

Skip to change the password and choose the Trident dashboard.