NFS in Kubernetes
Network File System is a ubiquitous technology that has applications even in today’s modern world of container clusters.
Using NFS is one of the easiest ways to get support for PVCs in Kubernetes that can dynamically provision their own PersistentVolumes.
Prerequisites
- A 1+ node Kubernetes cluster
- SSH in to each node and install the nfs-common OS package (e.g.
sudo apt-get update && sudo apt-get -y install nfs-common)
External Server Example
The NFS Client Provisioner can be used to leverage an existing NFS server to directly mount an existing NFS export into a pod.
The usual patterns should work here:
- Provision a VM outside of Kubernetes/Terraform (e.g. Ubuntu 16.04 LTS)
- SSH in and install the nfs-common OS package (e.g.
sudo apt-get update && sudo apt-get -y install nfs-common) - Create your
/exportsand run an NFS server - Open ports 2049, 20048, and 111 firewall using OpenStack security groups
- Consume the NFS mount from Kubernetes
Consuming the Mount
volumes:
- name: nfs
nfs:
server: <NFS_SERVER_IP>
path: /
Dynamic Volumes with the NFS Client Provisioner
For an easy way to get up and running serving shared volumes on Kubernetes from an existing NFS server, check out the nfs-client provisioner.
Provisioners like this one allow you to use Kubernetes to manage the lifecycle of your Kubernetes volumes and their data using PVCs, or PersistentVolumeClaims.
The basic workflow is as follows:
- User creates a PVC resource in Kubernetes with an attached StorageClass (if none is specified, the default will be used based on your cloud type - this blog post lists the default storage classes for each cloud provider)
- Kubernetes uses the StorageClass to determine how/whether a PV (PersistentVolume) should be created based on the claim parameters (e.g. for NFS, this effectively does a mkdir)
- If no existing static PV matches the parameters in the PVC, a new one should be dynamically created
- The PV can then be mounted into client pods by specifying the PVC name as a volume For more details, see Dynamic Volume Provisioning
In-Cluster Server Example
Following this example, I was able to easily get an NFS server pod running within a Kubernetes 1.9 cluster.
The original example was intended for use with GCE and included some nice features like backing your NFS server with PVCs.
Ideally, this could be handled by Terraform or some other deployment process, but this may change depending on the VM image specified (e.g. apt vs yum vs dnf vs apk, etc
Specific Modifications
The StorageClasses are specific to GCE, so we do not need to include those. Instead, we rely on our Terraform process mounting an external volume to the storage nodes.
Modifying the above example slightly to work with our Terraform/hostPath process, we have the following set of YAMLs:
kind: Service
apiVersion: v1
metadata:
name: nfs-server
spec:
ports:
- name: nfs
port: 2049
- name: mountd
port: 20048
- name: rpcbind
port: 111
selector:
role: nfs-server
---
apiVersion: v1
kind: ReplicationController
metadata:
name: nfs-server
spec:
replicas: 1
selector:
role: nfs-server
template:
metadata:
labels:
role: nfs-server
spec:
nodeSelector:
external-storage: "true"
containers:
- name: nfs-server
image: gcr.io/google_containers/volume-nfs:0.8
ports:
- name: nfs
containerPort: 2049
- name: mountd
containerPort: 20048
- name: rpcbind
containerPort: 111
securityContext:
privileged: true
volumeMounts:
- mountPath: /exports
name: nfs-export-fast
volumes:
- name: nfs-export-fast
hostPath:
path: /data/nfs
Consuming the Mount
The following example Pod consumes our in-cluster NFS export.
Note that the server field is populated with the Kubernetes Service IP (e.g. kubectl get svc):
apiVersion: v1
kind: Pod
metadata:
name: web
spec:
containers:
- name: web
image: nginx
volumeMounts:
# name must match the volume name below
- name: nfs
mountPath: "/usr/share/nginx/html/"
ports:
- name: web
containerPort: 80
protocol: TCP
volumes:
- name: nfs
nfs:
# FIXME: use the right name
#server: nfs-server.default.kube.local
server: "10.101.9.169"
path: "/"
readOnly: false
Testing it Out
After creating our NFS server and a pod consuming it, we can use kubectl exec to test that our NFS is working as expected:
# Check the node and name of our web pod
ubuntu@mltf-master:~$ kubectl get pods
NAME READY STATUS RESTARTS AGE IP NODE
nfs-server-wc8h6 1/1 Running 0 21m 10.244.1.4 mltf-storage0
web 1/1 Running 0 11m 10.244.3.2 mltf-storage1
# Exec into the container to test writing to the NFS
ubuntu@mltf-master:~$ kubectl exec -it web -- bash
root@web:/# cat /usr/share/nginx/html/index.html
Hello from NFS!
# Test writing to a file
root@web:/# vi /usr/share/nginx/html/pod-write
bash: vi: command not found
# No "vi" included in nginx image, so we install it
root@web:/# apt-get update && apt-get install vim
Ign:1 http://cdn-fastly.deb.debian.org/debian stretch InRelease
Get:2 http://security.debian.org/debian-security stretch/updates InRelease [94.3 kB]
... ... ... ... ... ... ... ... ...
Reading state information... Done
The following additional packages will be installed:
libgpm2 vim-common vim-runtime xxd
Suggested packages:
gpm ctags vim-doc vim-scripts
The following NEW packages will be installed:
libgpm2 vim vim-common vim-runtime xxd
0 upgraded, 5 newly installed, 0 to remove and 2 not upgraded.
Need to get 6766 kB of archives.
After this operation, 31.2 MB of additional disk space will be used.
Do you want to continue? [Y/n]
Get:1 http://cdn-fastly.deb.debian.org/debian stretch/main amd64 xxd amd64 2:8.0.0197-4+deb9u1 [132 kB]
Get:2 http://cdn-fastly.deb.debian.org/debian stretch/main amd64 vim-common all 2:8.0.0197-4+deb9u1 [159 kB]
... ... ... ... ... ... ... ... ...
update-alternatives: warning: skip creation of /usr/share/man/ja/man1/editor.1.gz because associated file /usr/share/man/ja/man1/vim.1.gz (of link group editor) doesn't exist
update-alternatives: warning: skip creation of /usr/share/man/man1/editor.1.gz because associated file /usr/share/man/man1/vim.1.gz (of link group editor) doesn't exist
# Test writing to a file
root@web:/# vi /usr/share/nginx/html/pod-write
root@web:/# cat /usr/share/nginx/html/pod-write
asdf 1234 Hello, World!
root@web:/# exit
exit
# SSH into "storage0", where our NFS server pod is running
ubuntu@mltf-master:~$ ssh 192.168.0.3
Welcome to Ubuntu 16.04.4 LTS (GNU/Linux 4.4.0-127-generic x86_64)
* Documentation: https://help.ubuntu.com
* Management: https://landscape.canonical.com
* Support: https://ubuntu.com/advantage
Get cloud support with Ubuntu Advantage Cloud Guest:
http://www.ubuntu.com/business/services/cloud
14 packages can be updated.
7 updates are security updates.
Last login: Mon Jun 11 16:28:49 2018 from 192.168.0.6
# Verify that the file created inside of our web pod (running on "storage1") was persisted to the NFS directory on "storage0":
ubuntu@mltf-storage0:~$ cat /data/nfs/pod-write
asdf 1234 Hello, World!
We can also ensure that these NFS volumes can be mounted into multiple pods simultaneously (e.g. ReadWriteMany):
# Create some duplicate pods consuming the same NFS mount
ubuntu@mltf-master:~/kubernetes-nfs-server$ kubectl create -f web-pod2.yaml -f web-pod3.yaml
pod "web2" created
pod "web3" created
# Wait for containers to start
ubuntu@mltf-master:~/kubernetes-nfs-server$ kubectl get pods -o wide
NAME READY STATUS RESTARTS AGE IP NODE
nfs-server-wc8h6 1/1 Running 0 49m 10.244.1.4 mltf-storage0
web 1/1 Running 0 39m 10.244.3.2 mltf-storage1
web2 0/1 ContainerCreating 0 8s <none> mltf-storage0
web3 0/1 ContainerCreating 0 8s <none> mltf-worker0
# Verify that the file created inside of our original web pod (running on "storage1") also shows up here
ubuntu@mltf-master:~/kubernetes-nfs-server$ kubectl exec -it web2 -- cat /usr/share/nginx/html/pod-write
asdf 1234 Hello, World!
ubuntu@mltf-master:~/kubernetes-nfs-server$ kubectl exec -it web3 -- cat /usr/share/nginx/html/pod-write
asdf 1234 Hello, World!
Dynamic Volumes using the NFS Provisioner
See NFS Provisioner
In a slightly more complex example, we can actually have Kubernetes provision volumes dynamically for us from an NFS export.
- Clone the repo above
- Locate the nfs/deploy/kubernetes subfolder
- Edit deployment.yaml (e.g. change nodeSelector, change mount path/config)
- kubectl create -f auth/serviceaccount.yaml -f auth/clusterrole.yaml -f auth/clusterrolebinding.yaml
- kubectl create -f class.yaml -f deployment.yaml
Congratulations! Wasn’t that easy?
Testing it Out
A simple test case for consuming a dynamic PVC can be found below:
kubectl create -f claim.yaml
kubectl create -f write_pod.yaml
Expected output:
ubuntu@mltf-master:~/external-storage/nfs/deploy/kubernetes$ kubectl get pvc,pv
NAME STATUS VOLUME CAPACITY ACCESS MODES STORAGECLASS AGE
persistentvolumeclaim/nfs Bound pvc-253140dc-6dc2-11e8-9b58-fa163eace583 1Mi RWX example-nfs 15m
NAME CAPACITY ACCESS MODES RECLAIM POLICY STATUS CLAIM STORAGECLASS REASON AGE
persistentvolume/pvc-253140dc-6dc2-11e8-9b58-fa163eace583 1Mi RWX Delete Bound default/nfs example-nfs 15m
# Look up the UID of our PVC (we will need this to find it on disk later)
ubuntu@mltf-master:~/external-storage/nfs/deploy/kubernetes$ kubectl get pvc -o yaml
apiVersion: v1
items:
- apiVersion: v1
kind: PersistentVolumeClaim
metadata:
annotations:
control-plane.alpha.kubernetes.io/leader: '{"holderIdentity":"c9839a10-6dc1-11e8-b790-0a580af40112","leaseDurationSeconds":15,"acquireTime":"2018-06-11T21:55:25Z","renewTime":"2018-06-11T21:55:27Z","leaderTransitions":0}'
... ... ... ... ... ... ... ... ... ...
volume.beta.kubernetes.io/storage-class: example-nfs
volume.beta.kubernetes.io/storage-provisioner: example.com/nfs
... ... ... ... ... ... ... ... ... ...
uid: 253140dc-6dc2-11e8-9b58-fa163eace583
spec:
... ... ... ... ... ... ... ... ... ...
requests:
storage: 1Mi
volumeName: pvc-253140dc-6dc2-11e8-9b58-fa163eace583
status:
... ... ... ... ... ... ... ... ... ...
# Wait for the write-pod to finish executing
ubuntu@mltf-master:~/external-storage/nfs/deploy/kubernetes$ kubectl get pods
NAME READY STATUS RESTARTS AGE NODE
mongo-k2pqv 1/1 Running 0 1h mltf-worker0
ndslabs-apiserver-pd62v 2/2 Running 0 1h mltf-storage1
ndslabs-etcd-ppsdv 1/1 Running 0 23m mltf-worker0
ndslabs-webui-qx4fb 1/1 Running 0 1h mltf-storage1
nfs-provisioner-7f78fcc699-vjr8d 1/1 Running 0 6m mltf-storage0
nfs-server-wc8h6 1/1 Running 0 5h mltf-storage0
web 1/1 Running 0 2h mltf-storage0
web2 1/1 Running 0 4h mltf-storage0
web3 1/1 Running 0 4h mltf-worker0
write-pod 0/1 Completed 0 2m mltf-storage1 <-------
# SSH into "storage0" (this is where the nfs-provisioner is currently running)
ubuntu@mltf-master:~/external-storage/nfs/deploy/kubernetes$ ssh 192.168.0.3
Welcome to Ubuntu 16.04.4 LTS (GNU/Linux 4.4.0-127-generic x86_64)
* Documentation: https://help.ubuntu.com
* Management: https://landscape.canonical.com
* Support: https://ubuntu.com/advantage
Get cloud support with Ubuntu Advantage Cloud Guest:
http://www.ubuntu.com/business/services/cloud
14 packages can be updated.
7 updates are security updates.
Last login: Mon Jun 11 21:55:53 2018 from 192.168.0.6
# Locate the subdirectory of the nfs-provisioner on our OpenStack volume for shared storage
ubuntu@mltf-storage0:~$ ls -al /data
total 32
drwxr-xr-x 5 root root 4096 Jun 11 21:51 .
drwxr-xr-x 26 root root 4096 Jun 11 21:19 ..
drwx------ 2 root root 16384 Jun 8 22:46 lost+found
drwxr-xr-x 4 root root 4096 Jun 11 20:52 nfs
drwxr-xr-x 3 root root 4096 Jun 11 21:55 nfs-provisioner
# Locate the subdirectory of our specific PVC (NOTE: this will match the "volume name" and UID of the PVC we looked up earlier)
ubuntu@mltf-storage0:~$ ls -al /data/nfs-provisioner/
total 36
drwxr-xr-x 3 root root 4096 Jun 11 21:55 .
drwxr-xr-x 5 root root 4096 Jun 11 21:51 ..
-rw-r--r-- 1 root root 14791 Jun 11 21:55 ganesha.log
-rw------- 1 root root 36 Jun 11 21:51 nfs-provisioner.identity
drwxrwsrwx 2 root root 4096 Jun 11 21:56 pvc-253140dc-6dc2-11e8-9b58-fa163eace583
-rw------- 1 root root 902 Jun 11 21:55 vfs.conf
# Verify the contents of the PVC
ubuntu@mltf-storage0:~$ ls -al /data/nfs-provisioner/pvc-253140dc-6dc2-11e8-9b58-fa163eace583/
total 8
drwxrwsrwx 2 root root 4096 Jun 11 21:56 .
drwxr-xr-x 3 root root 4096 Jun 11 21:55 ..
-rw-r--r-- 1 root root 0 Jun 11 21:56 SUCCESS <------