This page will walk you through the creation of a Presto Cluster in the Ahana Compute Plane in your AWS account.
It’s easy to get going. Let’s get started.
- Navigate to the Clusters tab
- Click on the Create Cluster button
Click on the icon to get additional help.
This needs to be unique across your Ahana compute plane. We recommend a descriptive name to help you identify clusters. It will also be used as part of the cluster endpoints. Example: https://telemetry.tenant.cp.ahana.cloud
The cluster name must begin and end with a letter or number. It can be a maximum of 63 characters long.
Select the Workload Profile based on the number of concurrent queries user expected to run on this cluster. There are two types of workload profiles.
Low Concurrency is useful for clusters that run a limited number of queries or a few large, complex queries. It also supports bigger and heavier ETL jobs.
High Concurrency is better for running multiple queries at the same time. For example, dashboard and reporting queries or A/B testing analytics, etc.
Concurrent queries are simply the number of queries executing at the same time in a given cluster. Hence, we have classified workloads based on the number of concurrent queries and curated a set of tuned session properties for each profile.
This setting can be changed once the cluster has been created, and a cluster restart is not required. However, the change will only apply to new queries.
Select the AWS EC2 instance type to be provisioned for the Presto cluster coordinator. Since Presto has only 1 coordinator, it is important to have an instance that can support the workload. Recommended instance type is r5.4xlarge
It is recommended to use a minimum AWS Instance type of
r5.4xlarge for the Presto coordinator and
r5.2xlarge for the Presto workers. You can find more information about the R5 instance class here.
The M5, R5, and C5 instance types will provide you with best price performance from the underlying Intel Cascade Lake Process Technology.
To learn more about Intel-optimized instances, visit the AWS and Intel Partner Page.
There are currently two types of scaling strategy you can use to manage your cluster:
- Scale Out only (CPU)
A Static scaling strategy means that the number of worker nodes is constant while the cluster is being used.
A Scale Out only (CPU) scaling strategy means that the number of worker nodes is increasing based on the worker nodes average CPU utilization.
Additionaly, both scaling strategies have the ability to scale in to the minimum number of worker nodes when the cluster is idle for a user-specified amount of time.
More details on the Presto Cluster Autoscaling page
Enter the number of worker nodes you want in the Presto cluster. Pick a number between 1 and 100. In addition to the workers, a coordinator is provisioned by default. The coordinator node is not included in this worker node count.
Manually reducing Presto workers on a cluster will gracefully shutdown workers so that any running queries will not fail due to the scale in. The query termination grace period is the maximum time window that is allowed for existing query tasks to complete on Presto workers before forcefully terminating those workers. The default is ten (10) minutes, and you can set a period between one (1) minute and two hours (120 minutes). You cannot edit this period once it is set. If you want a different query termination grace period in the future, you must create a new cluster.
You can optionally select to have Ahana create and managed a Hive Metastore for you for every cluster. Select the check box if you want the Ahana Hive Metastore to be provisioned.
If you select to create a Ahana-managed Hive Metastore, Ahana will provision a Hive Metastore named
ahana_hive that comes pre-configured and attached to the cluster. It also is pre-integrated with an S3 bucket. You can find information about the S3 bucket created on the cluster page. Use the
ahana_hive name in endpoints to connect to the HMS.
The Presto Query Log is always stored in an S3 bucket by default. It can also optionally be attached to the Presto Cluster to query it with Presto. By selecting this option, Ahana creates an external table and view in the attached Hive Metastore for easy access to the log.
Each Presto cluster must have at least one Presto user. If you already have a pool of created Presto users, you can simply select the Presto users you want to add to the cluster; you will be able to add or remove Presto users after the cluster is created as well.
If there are no Presto users or you do not see the Presto user you want, you can create a new Presto user by clicking on the Create Presto User button.
After you create the new Presto user, you will be able to add it to your cluster.
All Presto user authentication is done over HTTPS to secure your connection to clients, such the Presto CLI, JDBC drivers, and Superset.
Presto clusters created with compute plane versions below 3.0 only support a single Presto user per cluster. In this case, you will need to set a single set of credentials for the cluster by entering a username and password for the cluster.
If you are on a compute plane version below 3.0 and would like to be upgraded, please reach out to your Ahana representative or email
email@example.com. Once upgraded, any existing single user Presto clusters can be migrated to use multiple Presto users.
After you click the Create Cluster button, the cluster's page will initially show it in Pending state in the Pending Cluster table.
Once the cluster provisioning is complete, it will move into active state in the Active Clusters table.
Presto Clusters are provisioned on an Amazon EKS cluster that was initially created in your Ahana Compute Plane. Each node, coordinator and workers as well as the Hive Metastore are provisioned in individual instances. This means that there is only one container per instance in the Kubernetes cluster for complete isolation and resource utilization.