Achieving end to end Database management with PERCONA XtraDB cluster

Sigmoid
4 min readMay 16, 2023

--

Why Percona XtraDb Cluster on Kubernetes?

Databases on kubernetes are generally associated with several complexities and DataOps engineers are always on the look out for optimal solutions. Percona offers enterprise-grade database solutions, multiple replication, migration and support with the added advantage of being an open-source technology.

We leveraged Percona in our architectural model and this not only overcomes the issue of data losses with large data transactions working on the base of Mysql cluster solutions but also provides a high availability solution.

In this blog we are going to show an easy set up process to create a clustering solution using percona with K8s, Rancher and other open source tools. Whether it is fail-safe conditions,backup-issues or monitoring dashboards this set up provides enhanced results to optimize your cloud or on-prem environment.

Pre requisites:

Kubernetes cluster set up

Helm set up

Kubectl set up

Dive Into:

  • Run Kubectl get no (min 3 nodes cluster)

Set up Percoa-operator

We are using Percona Operator for MySQL based on Percona XtraDB Cluster. It is a Kubernetes operator to automatically manage failover scenarios.

Your app shouldn’t know anything about clustering, because it’s sorted out transparently under kubernetes-service-hostname:3306. So the app calls this address and behind it there are 3x SQLProxy/HAProxy containers (per server). Then the query is routed to one of three MySQL containers.

When the server goes down, failed SQLProxy/HAProxy and MySQL containers are removed from Kubernetes so kubernetes-service-hostname contains two instead of three members.

When the server is back online, containers are created to have a full cluster again.

There is also a Percona operator container which automatically helps to manage pods and do other actions so the cluster is fully operating.

In terms of storage, it can be just a hostPath local directory which shows a sign of simplicity from a storage perspective. You could also use PersistentVolumeClaim and any type of Storage Class behind it or external storage such as NFS.

  • Create namespace for percona in you cluster
Kubectl create namespace percona-sql
  • Add Percona repository via Helm
helm repo add percona https://percona.github.io/percona-
helm-charts/
helm repo update
  • Percona operators
helm upgrade --install pxc-operator percona/pxc-operator --
namespace percona-sql

Set up percona-sql on rke2

  • Create pv and storage class and pass storage class in values.yaml
Kubectl apply -f storagClass.yaml -n percona-sql

kind: StorageClass
apiVersion: storage.k8s.io/v1
metadata:
name: local-storage
provisioner: kubernetes.io/no-provisioner
volumeBindingMode: WaitForFirstConsumer
  • Create PV,PVC as per no of pods.
Kubectl apply -f percona-pv.yaml

apiVersion: v1
kind: PersistentVolume
metadata:
name: percona-pv
spec:
capacity:
storage: 10Gi
accessModes:
- ReadWriteOnce
persistentVolumeReclaimPolicy: Retain
storageClassName: local-storage-percona
local:
path: /mnt/percona/
nodeAffinity:
required:
nodeSelectorTerms:
- matchExpressions:
- key: kubernetes.io/hostname
operator: In
values:
- Node_Name
  • Install percona-sql via helm
helm upgrade --install pxc-db percona/pxc-db --set
haproxy.serviceType=NodePort,haproxy.replicasServiceType=No
dePort,pxc.persistence.enabled=true,pxc.persistence.storage
Class=local-
storage,pxc.persistence.accessMode=ReadWriteOnce,pxc.persis
tence.size=10Gi -n percona-sql
  • To access it via cli :
ROOT_PASSWORD=`kubectl -n percona-sql get secrets pxc-db -o 
jsonpath="{.data.root}" | base64 --decode`

kubectl run -i --tty --rm percona-client --image=percona --
restart=Never \ -- mysql -h pxc-db-haproxy.percona-
sql.svc.cluster.local -uroot -p"$ROOT_PASSWORD"

Percona Test cases: Fail-safe conditions for data loss

1. Pods restart

Steps

kubectl delete pods -n percona-sql –all
Kubectl get po -n percona-sql
kubectl run -i –tty –rm percona-client –
-image=percona –restart=Never — mysql -h pxc-db-haproxy.percona-
sql.svc.cluster.local -uroot -p”insecure-root-password”
validate your data
Results(Impact:Data Loss/No Data Loss)
Impact: data Loss
(No pv attached to pvc)

2. pv/pvc attach detach

Steps


Attach additional pv&pvc to the pxcdb and ha-proxy

vi percona-pv.yaml
apiVersion: v1
kind: PersistentVolume
metadata:
name: percona-pv
spec:
capacity:
storage: 10Gi
accessModes:
- ReadWriteOnce
persistentVolumeReclaimPolicy: Retain
storageClassName: local-storage-percona
local:
path: /mnt/percona/
nodeAffinity:
required:
nodeSelectorTerms:
- matchExpressions:
- key: kubernetes.io/hostname
operator: In
values:Node_Name
-
vi percona-pvc.yaml
apiVersion: v1
kind: PersistentVolumeClaim
metadata:
name: percona-pvc
namespace: percona-sql
spec:
accessModes:
- ReadWriteOnce
resources:
requests:
storage: 10Gi
storageClassName: local-storage-percona

Kubectl apply -f percona-pv.yaml -n percona-sql
Kubectl apply -f percona-pvc.yaml
-n percona-sql
Change values.yaml and then redeploy it.
Results(Impact:Data Loss/No Data Loss)
Impact:No data Loss
Description: Pvc/pv retains data

3. Node crash/Node detach

Each pv is attached to each node
If one node crash due to failure other pv’s can retain data.

Impact:No data Loss if pvc is attached

4. Pxc cluster delete/crash

Kubectl get pxc -n percona-sql

Impact: No data Loss if external pv (disk/local storage ) is attached even after pxc operator cluster deletion if remain (claim policy should be set to retain)

Conclusion:

We set up percona XtraDB on rke2 and tested it’s fail safe conditions that generally cause the problem of data loss. Preventing initial data loss rather than having a backup and restoring it was also one of the main highlights of this approach. Feel free to reach out to us at www.sigmoid.com/contact-us/ if you have any queries.

About the author

Akansha Immanuel is a DevOps Engineer at Sigmoid. She is a tech enthusiast and a voracious reader. She believes in dynamic progression in DevOps, and has also worked on several DataOps projects for large data producers and currently exploring FinOps.

--

--

Sigmoid

Innovative Data Solutions at Scale — Open Source | Cloud | ML | AI