Deploy Multiple Partitions

Performance of the Kubernetes environment can be improved by allocating multiple partitions for a Space.

The steps to allocate multiple partitions for a Space are shown below. This example allocates two partitions for the demo Space.

1. Install xap charts.
   

   helm install xap ./xap/ --version=15.2.0 --set manager.resources.limits.memory=1000Mi

2. Wait for all pods to be ready.

3. Install the demo Space with two partitions.

   helm install demo ./xap-pu/ --set manager.name=xap,partitions=2

4. Enter the command kubectl get pods to see the result:

NAME                            READY   STATUS    RESTARTS   AGE
demo-xap-pu-0                   1/1     Running   0          44s
demo-xap-pu-1                   1/1     Running   0          23s
xap-grafana-67d9b898b5-d5jn4    1/1     Running   0          2m27s
xap-influxdb-0                  1/1     Running   0          2m27s
xap-operator-55c7997cd5-rt6bt   1/1     Running   0          2m27s
xap-xap-manager-0               1/1     Running   0          2m26s

Deploy with High Availability

High availabilty deploys a backup partition that is in sync with the data in the primary partition. This eliminates downtime if the primary partition goes down, by seamlessly serving the same data from the backup.

When running in high availability mode, GigaSpaces combines backup partitions with anti-affinity rules to ensure that the space is intact in case a host or a zone (e.g. cloud availability zone) goes down. The GigaSpaces operator uses the anti-affinity pod level attribute to maintain high availability.

When a backup partition goes down, Kubernetes chooses to restart it on a new node according to the anti-affinity rules. In case of a primary partition malfunction, backup becomes primary and self healing is done in the same manner.

There are multiple options to define the topology of the anti-affinity rule. You can set up this option using the helm operator values.yaml file (under the antiAffinity.topology key):

• topology.kubernetes.io/zone (default) - new backups are restarted in a different zone than the primary

• topology.kubernetes.io/host - new backups are restarted in a different host (or virtual machine) than the primary.

• Any other topology defined by Kubernetes spec

How to set up a Cluster with Anti-affinity

Prerequisites

• A Kubernetes cluster with multiple zones (can be cloud installation)

• Gigaspaces helm charts xap and xap-pu pulled to a local repository (defined as TEST_REPO in this example)

Check environment settings

Verify that your cluster nodes are labeled by zone (default in most cloud environments):

kubectl get nodes --show-labels

Result:

NAME                            STATUS   ...   LABELS
ip-10-41-1-154.ec2.internal     Ready    ...   …,topology.kubernetes.io/zone=us-east-1a
ip-10-41-2-63.ec2.internal      Ready    ...   …,topology.kubernetes.io/zone=us-east-1b
ip-10-41-3-77.ec2.internal      Ready    ...   …,topology.kubernetes.io/zone=us-east-1c

Install xap umbrella chart

helm install xap TEST_REPO/xap/ --version=15.2.0 \
 --set operator.manager.name=xap \
 --set manager.metrics.enabled=false,metrics.enabled=false \
 --set manager.service.lrmi.enabled=false

Run the get pods command

 kubectl get pods

Result:

NAME                            READY   STATUS   RESTARTS    AGE
xap-operator-7b54b977f6-9znfd   1/1     Running   0          67s
ip-10-41-2-63.ec2.internal      1/1     Running   0          67s

Install a space with high availability and anti-affinity (by zone)

helm install demo TEST_REPO/xap/ --version=15.2.0 \
 --set manager.name=xap \
 --set partitions=2,ha=true,antiAffinity.enabled=true

Get pods and node allocation

kubectl get pods -o wide

Result:

NAME           ...      IP              NODE    
demo-xap-pu-0  ...      10.41.1.147     ip-10-41-1-154.ec2.internal [zone us-east-1a]
demo-xap-pu-1  ...      10.41.2.113     ip-10-41-2-63.ec2.internal  [zone us-east-1b]
demo-xap-pu-2  ...      10.41.3.142     ip-10-41-3-77.ec2.internal  [zone us-east-1c]
demo-xap-pu-3  ...      10.41.2.97      ip-10-41-2-63.ec2.internal  [zone us-east-1b]

Note that primary and backups are ordered in pairs and are in different zones:

• demo-xap-pu-0 and demo-xap-pu-1 are primary/backup of the first partition and are in zones 1a and 1b

• demo-xap-pu-2 and demo-xap-pu-3 are primary/backup of the second partition and are in zones 1c and 1b.