aerolab

Deploying clusters and clients with rack awareness

Deployment example description

• 2 zones/racks
• 2 aerospike nodes per rack - one per node
• 2 asbench client machines on a separate set of node (no load interference with servers)

Flow

To achieve this the following has to be met:

• the nodes need to be labelled, with an app= label to pick which nodes are used for aerospike and which can be used by asbench
• the kubectl yaml files must correctly select based on the labels for both asbench and the servers

Process

Deploy the EKS cluster with 2 nodegroups

cd /root/eks
eksctl create cluster -f two-node-groups.yaml

Deploy OLM and AKO

cd /root/deploy-olm-ako
./setup_olm.sh -f /root/eks/two-node-groups.yaml

Set labels

List nodes: kubectl get nodes -o wide

Simple script to mark first 6 nodes as aerospike app:

AEROSPIKE=6
kubectl get nodes -o wide --no-headers |awk '{print $1}' |sort |head -${AEROSPIKE} |while read node; do kubectl label nodes ${node} app=aerospike; done

View labels: kubectl get nodes --show-labels -o wide

Simple script to make the last 2 nodes astools:

ASTOOLS=2
kubectl get nodes -o wide --no-headers |awk '{print $1}' |sort |tail -${ASTOOLS} |while read node; do kubectl label nodes ${node} app=astools; done

Deploy aerospike cluster with label selector and 2 racks

cd /root/deploy-olm-ako/examples/clusters
kubectl apply -f aerospike_memory_rackaware_and_nodeselector.yaml -n aerospike

Deploy clients

cd /root/deploy-olm-ako/examples/clients
kubectl apply -f asbench-nodeselector-and-tolerations.yaml -n aerospike

This site is open source. Improve this page.