Workload orchestration: HashiCorp sends Nomad 0.12 across clusters

Nomad orchestration

DevOps tool provider HashiCorp’s application and container orchestrator Nomad has recently hit version 0.12, bringing multi-cluster deployments to enterprise customers among other things.

Though cluster federation has been available for a while, a new add-on module for Nomad Enterprise allows users to submit a job to multiple clusters without teams having to write their own automations. The fully supported enhancement also helps to coordinate job rollouts depending on the health of other deployments, and is meant to be especially beneficial for high availability scenarios.

In addition, enterprise users now have the option to query jobs and allocations across namespaces for easier debugging, and let their system automatically store snapshots of a cluster’s state as a backup mechanism.

Developers using the regular open source version of Nomad however were also given a few interesting new features to try. One of them is the so-called spread scheduling: The alternative to bin packing looks to let users evenly deploy applications across a cluster to distribute loads in a more economical manner. Nomad’s UI has become a bit more useful, presenting admins with ways to adjust the count of a task group, and stream client and agent logs for monitoring purposes.

Nomad 0.12 has also come loaded with the option to bind applications to specific network interfaces, and enable memory oversubscription for applications deployed with Docker task driver. Teams more keen on using Podman instead are free to do so thanks to a new driver for the Red Hat tool, while those looking to deploy and manage Windows-based web applications can do so now without containerising them first. A complete list of changes can be found in the release notes of the Nomad repository.

Along with the Nomad release, HashiCorp promoted the Nomad Autoscaler from preview into its beta phase, upping the tool’s version number to 0.1. The autoscaling daemon now supports horizontal autoscaling for both applications and clusters (for example on AWS), which is meant to help organisations in meeting their service-level agreements, and make sure scheduled applications get their resource needs met. Other new additions include support for autoscaling groups in AWS.