Setting up DBaaS¶

Do not use for mission-critical workloads

DBaaS feature is deprecated. We encourage you to use Percona Everest instead. Check our Migration guide.

To use the Database as a Service (DBaaS) solution in PMM there are a few things that need to be setup first including a suitable Kubernetes Cluster. If you’ve already got a kubernetes cluster you can jump ahead and enable DBaaS in PMM.

If you don’t have a Kubernetes cluster available you can use the free K8s provided by Percona for evaluation which will allow you to play around with DBaaS for 3 hours before the cluster expires. For a Kubernetes cluster that doesn’t expire you can use our “easy script”, you can find the instructions here.

In the sections that follow we’ll try to outline the steps to create your own Kubernetes cluster in a few popular ways.

Red Hat, CentOS¶

yum-config-manager --add-repo https://download.docker.com/linux/centos/docker-ce.repo
yum -y install docker-ce
usermod -a -G docker centos
systemctl enable docker
systemctl start docker

Debian, Ubuntu¶

apt-add-repository https://download.docker.com/linux/centos/docker-ce.repo
systemctl enable docker
systemctl start docker

minikube¶

Please follow minikube’s documentation to install it.

Red Hat, CentOS¶

yum -y install curl
curl -Lo /usr/local/sbin/minikube https://storage.googleapis.com/minikube/releases/latest/minikube-linux-amd64
chmod +x /usr/local/sbin/minikube
ln -s /usr/local/sbin/minikube /usr/sbin/minikube
alias kubectl='minikube kubectl --'

Start PMM server and activate a DBaaS feature¶

To start a fully-working 3 node XtraDB cluster, consisting of sets of 3x HAProxy, 3x PXC and 6x PMM Client containers, you will need at least 9 vCPU available for minikube. (1x vCPU for HAProxy and PXC and 0.5vCPU for each pmm-client containers).
DBaaS does not depend on PMM Client.
You can pass the environment variable --env ENABLE_DBAAS=1 to force the DBaaS feature when starting up pmm-server container. You can omit the variable and enable the feature later using PMM UI, please follow the link in step 3. below.
Add the option --network minikube if you run PMM Server and minikube in the same Docker instance. (This will share a single network and the kubeconfig will work.)
Add the options --env PMM_DEBUG=1 and/or --env PMM_TRACE=1 if you need extended debug details

Start PMM server:

docker run --detach --publish 80:80 --publish 443:443 --name pmm-server percona/pmm-server:2

Change the default administrator credentials:

(This step is optional, because the same can be done from the web interface of PMM on first login.)
```
docker exec -t pmm-server change-admin-password <new_password>
```

Important

You must activate DBaaS using the PMM UI if you omitted --env ENABLE_DBAAS=1 when starting up the container.

Create a Kubernetes cluster¶

The DBaaS feature uses Kubernetes clusters to deploy database clusters. You must first create a Kubernetes cluster and then add it to PMM using kubeconfig to get a successful setup.

Here are links to the current Kubernetes versions supported by DBaaS:

Minikube¶

Configure and start minikube:
```
minikube start --cpus=16 --memory=32G
```
Get your kubeconfig details from minikube. (You need these to register your Kubernetes cluster with PMM Server):
```
minikube kubectl -- config view --flatten --minify
```
You will need to copy this output to your clipboard and continue with adding a Kubernetes cluster to PMM.

Amazon AWS EKS¶

Create your cluster via eksctl or the Amazon AWS interface. For example:
```
eksctl create cluster --write-kubeconfig --name=your-cluster-name --zones=us-west-2a,us-west-2b --kubeconfig <PATH_TO_KUBECONFIG>
```
2. Copy the resulting kubeconfig and follow these instructions to register a Kubernetes cluster to PMM.

Google GKE¶

Create your cluster either with Google Cloud Console or gcloud command line tool:

The command below assumes that your gcloud command line tool is properly configured and your user authenticated and authorized to manage GKE Clusters. This example creates a minimal zonal cluster using preemptive node machines, ideal for testing the DBaaS functionality.

gcloud container clusters create --zone europe-west3-c pmm-dbaas-cluster --cluster-version 1.19 --machine-type e2-standard-4 --preemptible --num-nodes=3
gcloud container clusters get-credentials pmm-dbaas-cluster --zone=europe-west3-c
kubectl create clusterrolebinding cluster-admin-binding --clusterrole=cluster-admin --user=<<your_user@your_company.com>>

Create ServiceAccount, ClusterRole and RoleBindings (required Roles are deployed automatically when PMM deploys Operators) using the following command:

cat <<EOF | kubectl apply -f -
---
apiVersion: v1
kind: ServiceAccount
metadata:
  name: percona-dbaas-cluster-operator
---
kind: RoleBinding
apiVersion: rbac.authorization.k8s.io/v1beta1
metadata:
  name: service-account-percona-server-dbaas-xtradb-operator
subjects:
- kind: ServiceAccount
  name: percona-dbaas-cluster-operator
roleRef:
  kind: Role
  name: percona-xtradb-cluster-operator
  apiGroup: rbac.authorization.k8s.io
---
kind: RoleBinding
apiVersion: rbac.authorization.k8s.io/v1beta1
metadata:
  name: service-account-percona-server-dbaas-psmdb-operator
subjects:
- kind: ServiceAccount
  name: percona-dbaas-cluster-operator
roleRef:
  kind: Role
  name: percona-server-mongodb-operator
  apiGroup: rbac.authorization.k8s.io
---
apiVersion: rbac.authorization.k8s.io/v1beta1
kind: ClusterRole
metadata:
  name: service-account-percona-server-dbaas-admin
rules:
- apiGroups: ["*"]
  resources: ["*"]
  verbs: ["*"]
---
apiVersion: rbac.authorization.k8s.io/v1beta1
kind: ClusterRoleBinding
metadata:
  name: service-account-percona-server-dbaas-operator-admin
subjects:
- kind: ServiceAccount
  name: percona-dbaas-cluster-operator
  namespace: default
roleRef:
  kind: ClusterRole
  name: service-account-percona-server-dbaas-admin
  apiGroup: rbac.authorization.k8s.io
EOF

Extract variables required to generate a kubeconfig:

name=`kubectl get serviceAccounts percona-dbaas-cluster-operator -o json | jq  -r '.secrets[].name'`
certificate=`kubectl get secret $name -o json | jq -r  '.data."ca.crt"'`
token=`kubectl get secret $name -o json | jq -r  '.data.token' | base64 -d`
server=`kubectl cluster-info | grep 'Kubernetes control plane' | cut -d ' ' -f 7`

Generate your kubeconfig file (copy the output):

echo "
apiVersion: v1
kind: Config
users:
- name: percona-dbaas-cluster-operator
  user:
    token: $token
clusters:
- cluster:
    certificate-authority-data: $certificate
    server: $server
  name: self-hosted-cluster
contexts:
- context:
    cluster: self-hosted-cluster
    user: percona-dbaas-cluster-operator
  name: svcs-acct-context
current-context: svcs-acct-context
"

Follow the instructions on How to add a Kubernetes cluster with kubeconfig from the previous step.

Deleting clusters¶

If a Public Address is set in PMM Settings, for each DB cluster an API Key is created which can be found on the page /graph/org/apikeys. You should not delete them (for now, until issue PMM-8045 is fixed) – once a DB cluster is removed from DBaaS, the related API Key is also removed.

For example, if you only run eksctl delete cluster to delete an Amazon EKS cluster without cleaning up the cluster first, there will be a lot of orphaned resources such as Cloud Formations, Load Balancers, EC2 instances, Network interfaces, etc. The same applies for Google GKE clusters.

Cleaning up Kubernetes cluster¶

You should delete all database clusters, backups and restores.

kubectl delete perconaxtradbclusterbackups.pxc.percona.com --all
kubectl delete perconaxtradbclusters.pxc.percona.com --all
kubectl delete perconaxtradbclusterrestores.pxc.percona.com --all

kubectl delete perconaservermongodbbackups.psmdb.percona.com --all
kubectl delete perconaservermongodbs.psmdb.percona.com --all
kubectl delete perconaservermongodbrestores.psmdb.percona.com --all

In the dbaas-controller repository, in the deploy directory there are manifests we use to deploy operators. Use them to delete operators and related resources from the cluster.
Important
- Do NOT execute this step before all database clusters, backups and restores are deleted in the previous step. It may result in not being able to delete the namespace DBaaS lives in.
- Also be careful with this step if you are running DBaaS in more than one namespace as it deletes cluster level CustomResourceDefinitions needed to run DBaaS. This would break DBaaS in other namespaces. Delete just operators deployments in that case.
```
# Delete the PXC operator and related resources.
curl https://raw.githubusercontent.com/percona-platform/dbaas-controller/7a5fff023994cecf6bde15705365114004b50b41/deploy/pxc-operator.yaml | kubectl delete -f -

# Delete the PSMDB operator and related resources.
curl https://raw.githubusercontent.com/percona-platform/dbaas-controller/7a5fff023994cecf6bde15705365114004b50b41/deploy/psmdb-operator.yaml | kubectl delete -f -
```
Delete the name space where the DBaaS is running, this will delete all remaining name space level resources if any are left.
```
kubectl delete namespace <your-namespace>
```
Delete the Kubernetes cluster. The method depends on your cloud provider:
- Delete GKE cluster.
- Delete Amazon EKS cluster.

Run PMM Server as a Docker container for DBaaS¶

Start PMM server from a feature branch:
```
docker run --detach --name pmm-server --publish 80:80 --publish 443:443 --env ENABLE_DBAAS=1  percona/pmm-server:2;
```
Important
- Use --network minikube if running PMM Server and minikube in the same Docker instance. This way they will share single network and the kubeconfig will work.
- Use Docker variables --env PMM_DEBUG=1 --env PMM_TRACE=1 to see extended debug details.
Change the default administrator credentials:

This step is optional, because the same can be done from the web interface of PMM on the first login.
```
docker exec -t pmm-server change-admin-password <new_password>
```
Set the public address for PMM Server in PMM settings UI
Follow the steps for Add a Kubernetes cluster.
Follow the steps for Add a DB Cluster.
Get the IP address to connect your app/service:
```
minikube kubectl get services
```

Exposing PSMDB and XtraDB clusters for access by external clients¶

To make services visible externally, you create a LoadBalancer service or manually run commands to expose ports:

kubectl expose deployment hello-world --type=NodePort.

Get expert help¶

If you need assistance, visit the community forum for comprehensive and free database knowledge, or contact our Percona Database Experts for professional support and services.

Community Forum Get a Percona Expert