The Azure Disk CSI Driver V2 enhances the Azure Disk CSI Driver to improve scalability and reduce pod failover latency. It uses shared disks to provision attachment replicas on multiple cluster nodes and integrates with the pod scheduler to ensure a node with an attachment replica is chosen on pod failover. It is beneficial for both single zone use case as well as multi zone use case that uses Zone Redundant Disks. This demo is based on this guide
In this demo we will create a 3 node cluster distributed across 3 availability zones, deploy a single mysql pod, ingest some data there, and then drain the node hosting the pod. This means that we took one node offline, triggering the scheduler to migrate your pod to another node.
- Create the cluster
# Set the parameters
export LOCATION=northeurope # Location
export AKS_NAME=az-zrs
export RG=$AKS_NAME-$LOCATION
export AKS_CLUSTER_NAME=$AKS_NAME-cluster # name of the cluster
export K8S_VERSION=$(az aks get-versions -l $LOCATION --query 'orchestrators[-1].orchestratorVersion' -o tsv)
# Create RG
az group create --name $RG --location $LOCATION
# create the cluster
az aks create \
-g $RG \
-n $AKS_CLUSTER_NAME \
-l $LOCATION \
--kubernetes-version $K8S_VERSION \
--zones 1 2 3 \
--generate-ssh-keys
# get the credentials
az aks get-credentials -n $AKS_CLUSTER_NAME -g $RG
# verify access to the cluster
kubectl get nodes
NAME STATUS ROLES AGE VERSION
aks-nodepool1-20996793-vmss000000 Ready agent 77s v1.21.2
aks-nodepool1-20996793-vmss000001 Ready agent 72s v1.21.2
aks-nodepool1-20996793-vmss000002 Ready agent 79s v1.21.2- Install the Azure Disk CSI Driver V2
As of K8s 1.21, you can see the default storage class now pointing to the Azure Disk CSI Driver V1. To demonstrate the Azure Disk CSI Driver V2, we will install it side-by-side with the V1 driver. Tips for troubleshooting helm installation here.
helm repo add azuredisk-csi-driver https://raw.githubusercontent.com/kubernetes-sigs/azuredisk-csi-driver/master/charts
helm install azuredisk-csi-driver-v2 azuredisk-csi-driver/azuredisk-csi-driver \
--namespace kube-system \
--version v2.0.0-beta.2 \
--values=https://raw.githubusercontent.com/kubernetes-sigs/azuredisk-csi-driver/master/charts/v2.0.0-beta.2/azuredisk-csi-driver/side-by-side-values.yaml- Verify that the new storage classes were created
kubectl get storageclasses.storage.k8s.io
NAME PROVISIONER RECLAIMPOLICY VOLUMEBINDINGMODE ALLOWVOLUMEEXPANSION AGE
azuredisk-premium-ssd-lrs disk2.csi.azure.com Delete WaitForFirstConsumer true 2m20s
azuredisk-premium-ssd-zrs disk2.csi.azure.com Delete Immediate true 2m20s
azuredisk-standard-hdd-lrs disk2.csi.azure.com Delete WaitForFirstConsumer true 2m20s
azuredisk-standard-ssd-lrs disk2.csi.azure.com Delete WaitForFirstConsumer true 2m20s
azuredisk-standard-ssd-zrs disk2.csi.azure.com Delete Immediate true 2m20s
azurefile kubernetes.io/azure-file Delete Immediate true 2m30s
azurefile-csi file.csi.azure.com Delete Immediate true 2m30s
azurefile-csi-premium file.csi.azure.com Delete Immediate true 2m30s
azurefile-premium kubernetes.io/azure-file Delete Immediate true 2m30s
default (default) disk.csi.azure.com Delete WaitForFirstConsumer true 2m30s
managed kubernetes.io/azure-disk Delete WaitForFirstConsumer true 2m30s
managed-csi-premium disk.csi.azure.com Delete WaitForFirstConsumer true 2m30s
managed-premium kubernetes.io/azure-disk Delete WaitForFirstConsumer true 2m30sTo achieve faster pod failover and benefit from the replica mount feature of the Azure Disk CSI V2 driver, create a new storage class that sets the parameter for maxShares > 1.
apiVersion: storage.k8s.io/v1
kind: StorageClass
metadata:
name: azuredisk-standard-ssd-zrs-replicas
parameters:
cachingmode: None
skuName: StandardSSD_ZRS
maxShares: "2"
provisioner: disk2.csi.azure.com
reclaimPolicy: Delete
volumeBindingMode: Immediate
allowVolumeExpansion: true## Create ZRS storage class
kubectl apply -f zrs-replicas-storageclass.yaml
##validate that it was created
kubectl get sc | grep azuredisk- Create mysql statefulset using volumes provisioned by the Azure Disk CSI Driver V2 driver
- This deployment is based on this guide.
- The statefulset modified to use the azuredisk-standard-ssd-zrs-replicas StorageClass and Azure Disk CSI Driver V2 scheduler extender.
- The config map and service deployments can be taken straight from the guide.
Statefulset YAML Details
apiVersion: apps/v1
kind: StatefulSet
metadata:
name: mysql
spec:
selector:
matchLabels:
app: mysql
serviceName: mysql
replicas: 1
template:
metadata:
labels:
app: mysql
spec:
# Use the scheduler extender to ensure the pod is placed on a node with an attachment replica on failover.
schedulerName: csi-azuredisk-scheduler-extender
initContainers:
- name: init-mysql
image: mysql:5.7
command:
- bash
- "-c"
- |
set -ex
# Generate mysql server-id from pod ordinal index.
[[ `hostname` =~ -([0-9]+)$ ]] || exit 1
ordinal=${BASH_REMATCH[1]}
echo [mysqld] > /mnt/conf.d/server-id.cnf
# Add an offset to avoid reserved server-id=0 value.
echo server-id=$((100 + $ordinal)) >> /mnt/conf.d/server-id.cnf
# Copy appropriate conf.d files from config-map to emptyDir.
if [[ $ordinal -eq 0 ]]; then
cp /mnt/config-map/primary.cnf /mnt/conf.d/
else
cp /mnt/config-map/replica.cnf /mnt/conf.d/
fi
volumeMounts:
- name: conf
mountPath: /mnt/conf.d
- name: config-map
mountPath: /mnt/config-map
- name: clone-mysql
image: gcr.io/google-samples/xtrabackup:1.0
command:
- bash
- "-c"
- |
set -ex
# Skip the clone if data already exists.
[[ -d /var/lib/mysql/mysql ]] && exit 0
# Skip the clone on primary (ordinal index 0).
[[ `hostname` =~ -([0-9]+)$ ]] || exit 1
ordinal=${BASH_REMATCH[1]}
[[ $ordinal -eq 0 ]] && exit 0
# Clone data from previous peer.
ncat --recv-only mysql-$(($ordinal-1)).mysql 3307 | xbstream -x -C /var/lib/mysql
# Prepare the backup.
xtrabackup --prepare --target-dir=/var/lib/mysql
volumeMounts:
- name: data
mountPath: /var/lib/mysql
subPath: mysql
- name: conf
mountPath: /etc/mysql/conf.d
containers:
- name: mysql
image: mysql:5.7
env:
- name: MYSQL_ALLOW_EMPTY_PASSWORD
value: "1"
ports:
- name: mysql
containerPort: 3306
volumeMounts:
- name: data
mountPath: /var/lib/mysql
subPath: mysql
- name: conf
mountPath: /etc/mysql/conf.d
resources:
requests:
cpu: 500m
memory: 1Gi
livenessProbe:
exec:
command: ["mysqladmin", "ping"]
initialDelaySeconds: 30
periodSeconds: 10
timeoutSeconds: 5
readinessProbe:
exec:
# Check we can execute queries over TCP (skip-networking is off).
command: ["mysql", "-h", "127.0.0.1", "-e", "SELECT 1"]
initialDelaySeconds: 5
periodSeconds: 2
timeoutSeconds: 1
- name: xtrabackup
image: gcr.io/google-samples/xtrabackup:1.0
ports:
- name: xtrabackup
containerPort: 3307
command:
- bash
- "-c"
- |
set -ex
cd /var/lib/mysql
# Determine binlog position of cloned data, if any.
if [[ -f xtrabackup_slave_info && "x$(<xtrabackup_slave_info)" != "x" ]]; then
# XtraBackup already generated a partial "CHANGE MASTER TO" query
# because we're cloning from an existing replica. (Need to remove the tailing semicolon!)
cat xtrabackup_slave_info | sed -E 's/;$//g' > change_master_to.sql.in
# Ignore xtrabackup_binlog_info in this case (it's useless).
rm -f xtrabackup_slave_info xtrabackup_binlog_info
elif [[ -f xtrabackup_binlog_info ]]; then
# We're cloning directly from primary. Parse binlog position.
[[ `cat xtrabackup_binlog_info` =~ ^(.*?)[[:space:]]+(.*?)$ ]] || exit 1
rm -f xtrabackup_binlog_info xtrabackup_slave_info
echo "CHANGE MASTER TO MASTER_LOG_FILE='${BASH_REMATCH[1]}',\
MASTER_LOG_POS=${BASH_REMATCH[2]}" > change_master_to.sql.in
fi
# Check if we need to complete a clone by starting replication.
if [[ -f change_master_to.sql.in ]]; then
echo "Waiting for mysqld to be ready (accepting connections)"
until mysql -h 127.0.0.1 -e "SELECT 1"; do sleep 1; done
echo "Initializing replication from clone position"
mysql -h 127.0.0.1 \
-e "$(<change_master_to.sql.in), \
MASTER_HOST='mysql-0.mysql', \
MASTER_USER='root', \
MASTER_PASSWORD='', \
MASTER_CONNECT_RETRY=10; \
START SLAVE;" || exit 1
# In case of container restart, attempt this at-most-once.
mv change_master_to.sql.in change_master_to.sql.orig
fi
# Start a server to send backups when requested by peers.
exec ncat --listen --keep-open --send-only --max-conns=1 3307 -c \
"xtrabackup --backup --slave-info --stream=xbstream --host=127.0.0.1 --user=root"
volumeMounts:
- name: data
mountPath: /var/lib/mysql
subPath: mysql
- name: conf
mountPath: /etc/mysql/conf.d
resources:
requests:
cpu: 100m
memory: 100Mi
volumes:
- name: conf
emptyDir: {}
- name: config-map
configMap:
name: mysql
volumeClaimTemplates:
- metadata:
name: data
spec:
accessModes: ["ReadWriteOnce"]
storageClassName: azuredisk-standard-ssd-zrs-replicas
resources:
requests:
storage: 256Gi## create the configmap
kubectl apply -f mysql-configmap.yaml
## create the headless service
kubectl apply -f mysql-services.yaml
## create the statefulset
kubectl apply -f mysql-statefulset.yaml
## check that 2 services were created (headless one for the statefulset and mysql-read for the reads)
kubectl get svc -l app=mysql
NAME TYPE CLUSTER-IP EXTERNAL-IP PORT(S) AGE
mysql ClusterIP None <none> 3306/TCP 5h43m
mysql-read ClusterIP 10.0.205.191 <none> 3306/TCP 5h43m
## check the deployment (wait a bit until its running)
kubectl get pods -l app=mysql --watch
NAME READY STATUS RESTARTS AGE
mysql-0 2/2 Running 0 6m34s- now that the DB is running lets inject some data so we later can simulate failures
## create a database called v2test and a table called "messages", then inject a record in the database
kubectl run mysql-client --image=mysql:5.7 -i --rm --restart=Never --\
mysql -h mysql-0.mysql <<EOF
CREATE DATABASE v2test;
CREATE TABLE v2test.messages (message VARCHAR(250));
INSERT INTO v2test.messages VALUES ('Hello from V2');
EOF
If you don\'t see a command prompt, try pressing enter.
pod "mysql-client" deleted
## it may take a moment for the final delete to appear
## validate the data exist
kubectl run mysql-client --image=mysql:5.7 -i -t --rm --restart=Never --\
mysql -h mysql-read -e "SELECT * FROM v2test.messages"
+----------------+
| message |
+----------------+
| Hello from V2 |
+----------------+
pod "mysql-client" deleted- Simulate failure by cordoning and draining a node
## when we created our cluster we activated the availability zones feature, as we created 3 nodes, we should see that they are equally split across AZs
kubectl get nodes --output=custom-columns=NAME:.metadata.name,ZONE:".metadata.labels.topology\.kubernetes\.io/zone"
NAME ZONE
aks-nodepool1-20996793-vmss000000 northeurope-1
aks-nodepool1-20996793-vmss000001 northeurope-2
aks-nodepool1-20996793-vmss000002 northeurope-3
## lets check in which node our pods is running
kubectl get pods -l app=mysql -o wide
NAME READY STATUS RESTARTS AGE IP NODE NOMINATED NODE READINESS GATES
mysql-0 2/2 Running 0 17m 10.244.2.4 aks-nodepool1-20996793-vmss000001 <none> <none>
## We can see that the pod is running in "aks-nodepool1-20996793-vmss000001", through cordoning and draining the node it will trigger the pod to failover to a new node. So, lets try this out
kubectl cordon aks-nodepool1-20996793-vmss000001
node/aks-nodepool1-20996793-vmss000001 cordoned
kubectl drain aks-nodepool1-20996793-vmss000001
- At this moment our statefulset should try to restart in a different node in a new zone. With the Azure Disk CSI Driver V2, the pod should be up in just over a minute.
kubectl get pods -l app=mysql --watch -o wide
....
NAME READY STATUS RESTARTS AGE IP NODE NOMINATED NODE READINESS GATES
mysql-0 2/2 Running 0 10m 10.244.0.7 aks-nodepool1-20996793-vmss000002 <none> <none>
## Now that the pod failover is complete, lets validate that the client can access the server. We should see the data we wrote earlier
kubectl run mysql-client --image=mysql:5.7 -i -t --rm --restart=Never --\
mysql -h mysql-read -e "SELECT * FROM v2test.messages"
+----------------+
| message |
+----------------+
| Hello from V2 |
+----------------+
pod "mysql-client" deleted
## This showcases the speed at which the Azure Disk CSI Driver V2 can facilitate pod failover