Test Deployment Notes

2025-05-11

What follows is my, potentially slightly incoherent, stream of consciousness as I try to teach myself Kubernetes

First Deployment

Minikube running with QEMU driver on local machine
used the following deployment and service YAML:

 1---
 2apiVersion: apps/v1
 3kind: Deployment
 4metadata:
 5  name: uptime-kuma
 6  labels:
 7    app: uptime-kuma
 8spec:
 9  replicas: 1
10  selector:
11    matchLabels:
12      app: uptime-kuma
13  template:
14    metadata:
15      labels:
16        app: uptime-kuma
17    spec:
18      containers:
19      - name: uptime-kuma
20        image: louislam/uptime-kuma
21        ports:
22        - containerPort: 3001
23---
24apiVersion: v1
25kind: Service
26metadata:
27  name: uptime-kuma-service
28spec:
29  selector:
30    app: uptime-kuma
31  type: LoadBalancer  
32  ports:
33    - protocol: TCP
34      port: 3001
35      targetPort: 3001
36      nodePort: 30000

kubectl apply -f uptime-kuma-deployment.yml
Got “MK_UNIMPLEMENTED” error when running minikube service uptime-kuma-service to get URL. Google showed that minikube service was unavailable when using QEMU. Saw an option to use Docker instead.
ran minikube delete and redeployed minikube with minikube start --driver=docker
re-ran kubectl apply
minikube service command gave URL http://192.168.49.2:30000 which brought up uptime-kuma in the browser

Adding Volumes for Persistent Storage

 1---
 2apiVersion: apps/v1
 3kind: Deployment
 4metadata:
 5  name: uptime-kuma
 6  labels:
 7    app: uptime-kuma
 8spec:
 9  replicas: 1
10  selector:
11    matchLabels:
12      app: uptime-kuma
13  template:
14    metadata:
15      labels:
16        app: uptime-kuma
17    spec:
18      containers:
19      - name: uptime-kuma
20        image: louislam/uptime-kuma
21        ports:
22        - containerPort: 3001
23        volumeMounts: # Volume must be created along with volumeMount (see next below)
24        - name: uptime-kuma-data
25          mountPath: /app/data # Path within the container, like the right side of a docker bind mount -- /tmp/data:/app/data
26      volumes: # Defines a volume that uses an existing PVC (defined below)
27      - name: uptime-kuma-data
28        persistentVolumeClaim:
29          claimName: uptime-kuma-pvc
30---
31apiVersion: v1
32kind: Service
33metadata:
34  name: uptime-kuma-service
35spec:
36  selector:
37    app: uptime-kuma
38  type: LoadBalancer  
39  ports:
40    - protocol: TCP
41      port: 3001
42      targetPort: 3001
43      nodePort: 30000
44---
45apiVersion: v1
46kind: PersistentVolume
47metadata:
48  name: uptime-kuma-pv
49spec:
50  accessModes:
51    - ReadWriteOnce
52  capacity:
53    storage: 1Gi
54  storageClassName: standard
55  hostPath:
56    path: /tmp/uptime/data # Path on the host
57---
58apiVersion: v1
59kind: PersistentVolumeClaim
60metadata:
61  name: uptime-kuma-pvc
62spec:
63  accessModes:
64    - ReadWriteOnce
65  resources:
66    requests:
67      storage: 1Gi
68  storageClassName: standard
69  volumeName: uptime-kuma-pv

Was confused on the need for storageClassName – tried to comment it out and redeploy:
- Pod was stuck in pending status
- kubectl describe pod showed:Events:

  Type     Reason            Age    From               Message
  ----     ------            ----   ----               -------
  Warning  FailedScheduling  4m10s  default-scheduler  0/1 nodes are available: persistentvolumeclaim "uptime-kuma-pvc" not found. preemption: 0/1 nodes are available: 1 Preemption is not helpful for scheduling.
  Warning  FailedScheduling  4m9s   default-scheduler  0/1 nodes are available: pod has unbound immediate PersistentVolumeClaims. preemption: 0/1 nodes are available: 1 Preemption is not helpful for scheduling.

kubectl describe pvc uptime-kuma-pvc showed:

Events:
  Type     Reason          Age                  From                         Message
  ----     ------          ----                 ----                         -------
  Warning  VolumeMismatch  7s (x26 over 6m20s)  persistentvolume-controller  Cannot bind to requested volume "uptime-kuma-pv": storageClassName does not match

-kubectl describe pv uptime-kuma-pv shows that uptime-kuma-pv has no storage class assigned:

Name:            uptime-kuma-pv
Labels:          <none>
Annotations:     <none>
Finalizers:      [kubernetes.io/pv-protection]
StorageClass:    
Status:          Available

The docs state the following: You can create a PersistentVolumeClaim without specifying a storageClassName for the new PVC, and you can do so even when no default StorageClass exists in your cluster. In this case, the new PVC creates as you defined it, and the storageClassName of that PVC remains unset until a default becomes available.
Checking what my default storage class is with kubectl get storageclass (I didn’t set this… must be a minikube default)

NAME                 PROVISIONER                RECLAIMPOLICY   VOLUMEBINDINGMODE   ALLOWVOLUMEEXPANSION   AGE
standard (default)   k8s.io/minikube-hostpath   Delete          Immediate           false                  2d

It appears that a PVC will be set to the default storage class if one is not specified, but the same is not true for a PV – the docs also state: A PV with no storageClassName has no class and can only be bound to PVCs that request no particular class.
After understanding this I ran kubectl apply -f uptime-kuma-deployment.yml with the above configuration.
Ran minikube service uptime-kuma-service to pull it up in the browser
Created an uptime-kuma user and created a monitor for google.com
Ran kubectl delete deployment uptime-kuma to delete all pods, then re-ran kubectl apply -f uptime-kuma-deployment, then opened it in the browser again with minikube service uptime-kuma-service
My user and monitor still exist after recreating the deployment, so data was successfully persisted with a PV and PVC! 🎉🎉🎉

Config files are in the folder remote-storage-smb-deployment
Created a share on TrueNAS called ‘kubernetes-testing’
Installed the SMB CSI driver – https://github.com/kubernetes-csi/csi-driver-smb/tree/master
Generate a secret to store share credentials with kubectl create secret generic smbcreds --from-literal username=USERNAME --from-literal password="PASSWORD"
- note: the generic flag is used to denote that the secret being created is an ‘Opaque’ secret (arbitrary, user-defined data). There are other secret types for various specific use cases such as docker registry credentials or tls certificates. https://kubernetes.io/docs/concepts/configuration/secret/
Used this video from the Youtube channel Jim’s garage for reference with the PV and PVC configs https://www.youtube.com/watch?v=3S5oeB2qhyg&t=318s
Got a permission denied error on the pod, appears to be when mounting smb share in fstab
Ran the following to verify contents of the secret: kubectl get secret smbcreds -o jsonpath='{.data}' which gave the following output: {"password":"MXFhQFdTM2Vk","username":"a3VzZXI="}
These values are base64 encoded, so I ran the password through an online decoder and found that it appears to have been cut off after 9 characters.
Found out the reason that it was cut off was that when using double quotes the shell will interpret special characters. Since the password was a waterfall of 1qa@WS3ed$RF5tg –> the shell interpreted everything after the dollar sign as a variable (which was obviously empty)
After fixing this, everything appears to deploy without errors, however connecting with minikube service gives ‘ERR_CONNECTION_REFUSED’
Checking pod logs – it appears the app can’t connect to SQLLite because the database is locked
Switched to an nginx image rather than uptme-kuma – you wouldn’t want a SQLLite DB on an SMB share anyway.
Gave a mountpath of /usr/share/nginx/html
I thought if nothing existed for this volume it would create a default splashpage – it didn’t, and I got 403 forbidden
Added a hello-world HTML file to the share and then ran kubectl rollout restart deployment nginx to restart the pod(s)
It worked! :tada:
Scaled pods up from 1 to 2 and re-ran kubectl apply – still works! :tada:
Data is thrown into the share like data soup (no sub-directories based on pod, pvc etc)
Added this storage class config based on the docs:

 1apiVersion: storage.k8s.io/v1
 2kind: StorageClass
 3metadata:
 4  name: smb
 5provisioner: smb.csi.k8s.io
 6parameters:
 7  source: //10.0.30.11/kubernetes-testing
 8  # if csi.storage.k8s.io/provisioner-secret is provided, will create a sub directory
 9  # with PV name under source
10  csi.storage.k8s.io/provisioner-secret-name: smbcreds
11  csi.storage.k8s.io/provisioner-secret-namespace: default
12  csi.storage.k8s.io/node-stage-secret-name: smbcreds
13  csi.storage.k8s.io/node-stage-secret-namespace: default
14reclaimPolicy: Delete  # available values: Delete, Retain
15volumeBindingMode: Immediate
16allowVolumeExpansion: true
17mountOptions:
18  - dir_mode=0777
19  - file_mode=0777
20  - uid=1000
21  - gid=1000

Also removed the PersistentVolume config – this dynamically provisioned volumes and created sub-directories on the share based on the names of the dynamically provisioned volumes (ex. pvc-646faec2-c9cc-4a75-9b3a-4b6c9742f339)
I wanted a sub-directory created based on the deployment name rather than the difficult-to-remember PVC names.
I found two ways to do this:
- Add subDir parameter to the Storage Class config. This will create a single sub-directory that every PVC will use rather than a unique folder based on the PVC name.
- Add subPath in the Container Template config under volumeMounts – this will put the contents of that volume underneath that sub-folder
- Both of these methods can be used at the same time
Added both of these configurations, attempted to add additional heimdall deployment – same issue with database locking. Decided to try to deploy it as a statefulset:
When deleting heimdall’s PVC it deleted the entire subDir defined in the storage path
Found out the reason for this is that the default of the onDelte parameter is to delete the directories when the volume is deleted. Because the storageclass’s reclaim policy was Delete – the volume was deleted when I deleted the PVC. Therefore, the CSI driver deleted the directories on the share as well.
https://github.com/kubernetes-csi/csi-driver-smb/blob/master/docs/driver-parameters.md

Deploying a StatefulSet

#kubernetes

Test Deployment Notes

First Deployment

Adding Volumes for Persistent Storage

Using an SMB share as persistent storage

Deploying a StatefulSet