I am trying to run a docker container registry in Minikube for testing a CSI driver that I am writing.
I am running minikube on mac and am trying to use the following minikube start command: minikube start --vm-driver=hyperkit --disk-size=40g. I have tried with both kubeadm and localkube bootstrappers and with the virtualbox vm-driver.
This is the resource definition I am using for the registry pod deployment.
---
apiVersion: v1
kind: Pod
metadata:
name: registry
labels:
app: registry
namespace: docker-registry
spec:
containers:
- name: registry
image: registry:2
imagePullPolicy: Always
ports:
- containerPort: 5000
volumeMounts:
- mountPath: /var/lib/registry
name: registry-data
volumes:
- hostPath:
path: /var/lib/kubelet/plugins/csi-registry
type: DirectoryOrCreate
name: registry-dataI attempt to create it using kubectl apply -f registry-setup.yaml. Before running this my minikube cluster reports itself as ready and with all the normal minikube containers running.
However, this fails to run and upon running kubectl describe pod, I see the following message:
Name: registry
Namespace: docker-registry
Node: minikube/192.168.64.43
Start Time: Wed, 08 Aug 2018 12:24:27 -0700
Labels: app=registry
Annotations: kubectl.kubernetes.io/last-applied-configuration={"apiVersion":"v1","kind":"Pod","metadata":{"annotations":{},"labels":{"app":"registry"},"name":"registry","namespace":"docker-registry"},"spec":{"cont...
Status: Running
IP: 172.17.0.2
Containers:
registry:
Container ID: docker://42e5193ac563c2b2e2a2b381c91350d30f7e7c5009a30a5977d33b403a374e7f
Image: registry:2
...
TRUNCATED FOR SPACE
...
Events:
Type Reason Age From Message
---- ------ ---- ---- -------
Normal Scheduled 1m default-scheduler Successfully assigned registry to minikube
Normal SuccessfulMountVolume 1m kubelet, minikube MountVolume.SetUp succeeded for volume "registry-data"
Normal SuccessfulMountVolume 1m kubelet, minikube MountVolume.SetUp succeeded for volume "default-token-kq5mq"
Normal Pulling 1m kubelet, minikube pulling image "registry:2"
Normal Pulled 1m kubelet, minikube Successfully pulled image "registry:2"
Normal Created 1m kubelet, minikube Created container
Normal Started 1m kubelet, minikube Started container
...
TRUNCATED
...
Name: storage-provisioner
Namespace: kube-system
Node: minikube/192.168.64.43
Start Time: Wed, 08 Aug 2018 12:24:38 -0700
Labels: addonmanager.kubernetes.io/mode=Reconcile
integration-test=storage-provisioner
Annotations: kubectl.kubernetes.io/last-applied-configuration={"apiVersion":"v1","kind":"Pod","metadata":{"annotations":{},"labels":{"addonmanager.kubernetes.io/mode":"Reconcile","integration-test":"storage-provis...
Status: Pending
IP: 192.168.64.43
Containers:
storage-provisioner:
Container ID:
Image: gcr.io/k8s-minikube/storage-provisioner:v1.8.1
Image ID:
Port: <none>
Host Port: <none>
Command:
/storage-provisioner
State: Waiting
Reason: ErrImagePull
Ready: False
Restart Count: 0
Environment: <none>
Mounts:
/tmp from tmp (rw)
/var/run/secrets/kubernetes.io/serviceaccount from storage-provisioner-token-sb5hz (ro)
Conditions:
Type Status
Initialized True
Ready False
PodScheduled True
Volumes:
tmp:
Type: HostPath (bare host directory volume)
Path: /tmp
HostPathType: Directory
storage-provisioner-token-sb5hz:
Type: Secret (a volume populated by a Secret)
SecretName: storage-provisioner-token-sb5hz
Optional: false
QoS Class: BestEffort
Node-Selectors: <none>
Tolerations: node.kubernetes.io/not-ready:NoExecute for 300s
node.kubernetes.io/unreachable:NoExecute for 300s
Events:
Type Reason Age From Message
---- ------ ---- ---- -------
Normal Scheduled 1m default-scheduler Successfully assigned storage-provisioner to minikube
Normal SuccessfulMountVolume 1m kubelet, minikube MountVolume.SetUp succeeded for volume "tmp"
Normal SuccessfulMountVolume 1m kubelet, minikube MountVolume.SetUp succeeded for volume "storage-provisioner-token-sb5hz"
Normal Pulling 23s (x3 over 1m) kubelet, minikube pulling image "gcr.io/k8s-minikube/storage-provisioner:v1.8.1"
Warning Failed 21s (x3 over 1m) kubelet, minikube Failed to pull image "gcr.io/k8s-minikube/storage-provisioner:v1.8.1": rpc error: code = Unknown desc = failed to register layer: Error processing tar file(exit status 1): write /storage-provisioner: no space left on device
Warning Failed 21s (x3 over 1m) kubelet, minikube Error: ErrImagePull
Normal BackOff 7s (x3 over 1m) kubelet, minikube Back-off pulling image "gcr.io/k8s-minikube/storage-provisioner:v1.8.1"
Warning Failed 7s (x3 over 1m) kubelet, minikube Error: ImagePullBackOff
------------------------------------------------------------
...
So while the registry container starts up correctly, a few of the other minikube services (including dns, http ingress service, etc) begin to fail with reasons such as the following: write /storage-provisioner: no space left on device. Despite allocating a 40GB disk-size to minikube, it seems as though minikube is trying to write to rootfs or devtempfs (depending on the vm-driver) which has only 1GB of space.
$ df -h
Filesystem Size Used Avail Use% Mounted on
rootfs 919M 713M 206M 78% /
devtmpfs 919M 0 919M 0% /dev
tmpfs 996M 0 996M 0% /dev/shm
tmpfs 996M 8.9M 987M 1% /run
tmpfs 996M 0 996M 0% /sys/fs/cgroup
tmpfs 996M 8.0K 996M 1% /tmp
/dev/sda1 34G 1.3G 30G 4% /mnt/sda1Is there a way to make minikube actually use the 34GB of space that was allocated to /mnt/sda1 instead of rootfs when pulling images and creating containers?
Thanks in advance for any help!
You can also use the minikube --docker-opt option to set the --data-root option of the dockerd daemon running inside minikube. --docker-opt can be used as a pass-through for any parameter to dockerd.
For example, in the case you describe above it would look like:
minikube start --vm-driver=hyperkit --disk-size=40g --docker-opt="--data-root /mnt/sda1"
Keep in mind that if you try to modify an existing minikube cluster you either have to copy var/lib/docker to /mnt/sda1 (as the previous answer also suggested) before restarting or delete and rebuild the cluster.
update: After experimentation, I noticed that the above solution will not work the first time you run minikube start as it somehow interferes with minikube's own core-system build and boot-up process. In practice this means that you need to run minikube start at least once without the --docker-opt to build the core system and then re-run it with --docker-opt.
You need to configure your Minikube virtual machine for using /dev/sda1 instead of / for Docker. To log in to it, use minikube ssh command.
Than you have two options:
Mount /dev/sda1 to var/lib/docker, but don't forget to copy the content from original var/lib/docker to /mnt/sda1 before that.
Reconfigure Docker for using /mnt/sda1 instead of var/lib/docker for storing images. Look through this link for more information about it.