Facilitating the Spread of Knowledge and Innovation in Professional Software Development

Write for InfoQ


Choose your language

InfoQ Homepage News Azure Brings Vertical Scaling, Monitor Alerts and More for Apache Cassandra Managed Instance

Azure Brings Vertical Scaling, Monitor Alerts and More for Apache Cassandra Managed Instance

This item in japanese

Microsoft has recently released several new features for Azure Managed Instance for Apache Cassandra, such as upgrading the Apache Cassandra version to 4.0 GA, Azure Monitor alerts and insights, deallocating cluster resources to improve costs, vertical scaling and more.

First released in 2021 by the Azure Cosmos DB engineers, Azure Managed Instance for Apache Cassandra was designed to simplify the life of developers by providing automated deployment and scaling operations for managed open-source Apache Cassandra datacenters with the scaling up and scaling down nodes fully managed by the orchestrator within the Cassandra ring. Other Apache Cassandra tasks that the service manages automatically are provisioning a cluster, starting a repair action on a keyspace, setting up backups and maintaining audit logs.

Apache Cassandra 4.0 GA is the default version when deploying clusters using the Azure portal or CLI; some of the features added to version 4.0 GA include support for Java 11, virtual tables, full query logging, audit logging, streaming and more.

It is also now possible to modify the Cassandra nodes by scaling vertically and horizontally. In order to scale up to a more powerful SKU, you only need to select from the Sku size dropdown, and then click in Scale. It is important to note that this operation can take several minutes, which means that when Azure informs the operation is complete, this doesn't mean that all the nodes have joined the Cassandra ring yet.

(image taken from

Aiming to optimize costs, Azure Managed Instance for Apache Cassandra allows an operator to deallocate/pause resources in a non-production cluster in order to avoid being charged; it is important to notice that storage is still being charged for deallocated resources. To deallocate, first change the cluster type for NonProduction and then deallocate.

(image taken from

Azure Monitor now brings alerts for Azure Managed Instance for Apache Cassandra; alerts are designed to help to detect and address issues in the infrastructure or application before the users notice them. It is possible to configure alerts on any metric or log data in the Azure Monitor data platform. Let's see some types of alerts available:

  • Metric alerts - platform, custom, application insights metrics and logs converted to metrics evaluated at regular intervals
  • Log alerts - resource logs evaluated through Log Analytics query at a predefined frequency
  • Prometheus alerts - alerts based on Prometheus metrics stored in Azure Monitor managed services for Prometheus

The complete list of types of alerts can be found here.

Azure Monitor Insights or curated visualizations are available in preview mode for Azure Managed Instance for Apache Cassandra. This consists of a subset of the complete metrics to monitor the clusters and provide visualizations. Following are some of the curated visualizations:

  • Azure VM Insights - analyze performance and health of Azure VMs and Virtual Machine Scale Sets at scale
  • Azure Container Insights - monitor the performance of container workloads deployed to managed Kubernetes clusters hosted on Azure Kubernetes Service, giving performance visibility by collecting metrics from controllers, nodes and containers

The complete list of Insights is available here.

Cassandra Lucene Index, derived from Stratio Cassandra, is a plugin that extends its index functionality to provide full-text search capabilities and free bitemporal, multivariable, and geospatial search. Through an Apache Lucene-based implementation of Cassandra secondary indexes, each node of the clusters indexes its own data. It is important to note that the Lucene Index feature is in public preview and not recommended for production workloads.

About the Author

Rate this Article