The repo contains scripts that create a scylla cluster on EKS.
Creating the cluster:
0) Create a file cluster/terraform.tfvars
(see the terraform.tfvars.example
).
Specify the cluster name, region, vpc id and the private vpc networks.
- Create the cluster
make create_cluster
- (this step will be automated) Create the daemonset that mounts the instance store to the ec2 instances:
kubectl --kubeconfig=cluster/kubeconfig_{cluster_name} apply -f node-setup.yaml
There will be an EKS cluster created with a single-az Scylla cluster with 2 nodes.
- connecting to the scylla cluster from outside the cluster
- multi-az setup
- enable cpu pinning, node networking
- scylla monitoring
- running a cassandra stress test
- backup and restoring a new cluster from backup (scylla manager, velero, a cross-region standby node?)
- testing the failover capabilities: 1 node is removed, 2 nodes removed. Are ip addresses kept?
- what is the "repair" operation: is it needed, how to run it?
- how to upgrade the cluster
- scylla monitoring loki
- how Spark/Athena read from scylla? How to dump the current state into s3 in Parquet?