The future of orchestration is heterogeneous. Platform teams should not have to choose between containers, raw binaries, and virtual machines, or maintain separate, siloed control planes for each. Today’s infrastructure demands an orchestration platform that pairs high-throughput scheduling with granular kernel-level security and clean operational automation across them.
This year, engineering across the Nomad ecosystem has focused on advancing that foundation. Spanning the Nomad 2.0 and 2.1 release cycles, from core consensus and performance enhancements to new scheduling governance, runtime drivers, autoscaling, and Terraform tooling, Nomad continues to evolve to meet the demands of modern production environments.
»Improve control plane performance and recovery
Nomad 2.0 lets operators migrate the Raft log store from BoltDB to a Write-Ahead Log (WAL) backend, which offers better performance for write-heavy workloads. Nomad 2.0.5 added the plan_apply_pipeline setting, which lets operators control how many outstanding plans the leader allows when evaluating plans.
For recovery, Nomad 2.1 adds nomad operator snapshot filter. It removes selected entries from a snapshot and writes the result to a copy. Operators can use it to clean up a snapshot before restoring it, or to remove a large number of pending evaluations when recovering a cluster from an outage.
»Govern high-volume batch bursts with node pool queueing
High-volume batch bursts, like automated pipelines or ML inference, can leave smaller jobs stuck behind larger ones when capacity runs short. Nomad 2.1 introduces the experimental Batch Job Queue to give operators greater control over how batch jobs are scheduled within node pools.
The queue holds pending batch jobs and provides two queueing options, FIFO, and a dynamic priority implementation which adjusts the priority of queued jobs based on multiple configurable factors. This helps prevent tenant starvation and supports fair-share scheduling.
»Advance containerless security and storage with Driver exec2
exec2 v0.2.0 brings container-grade capabilities to non-containerized workloads, adding Linux ambient capabilities, Host and CSI volume mounts, and nomad alloc exec support without container daemon overhead.
The exec2 task driver leverages native Linux kernel namespaces and Landlock filesystem sandboxing to isolate raw binary tasks. With v0.2.0, operators can selectively grant Linux ambient capabilities using allow_caps, cap_add, and cap_drop — a web service can bind privileged ports via CAP_NET_BIND_SERVICE while dropping all other root privileges. exec2 tasks can now mount persistent storage via standard Nomad Host and CSI volume blocks, native nomad alloc exec support enables interactive shell debugging and script-based health checks, and memory telemetry transitions to modern Linux standards, accurately reporting Resident Set Size (RSS) under cgroups v2.
»Expand workload diversity with rootless Podman and virtual machines
Orchestration flexibility extends beyond containers and binaries to include rootless environments and virtual machines:
Rootless Podman (v0.6.5): Nomad Driver Podman 0.6.5 streamlines its dependency footprint by removing transitive Docker Go packages while adding multi-architecture platform overrides (os, arch, variant), direct HTTP(S) image archive loading, and shared IPC namespaces for inter-task /dev/shm communication.
Virtual Machine Orchestration with Virt (Tech Preview): The Virt Task Driver v0.0.2-beta.1 (Tech Preview) advances native virtual machine orchestration on Nomad, adding support for host directory and Ceph storage pools, block-device volume attachments, and macvtap network interfaces.
»Strengthen infrastructure resilience with APM-driven autoscaling and fail-safe drains
Effective autoscaling requires both application-aware metrics and fail-safe teardowns. Relying solely on CPU or memory thresholds often lags behind real traffic surges, while aggressive node scale-in risks terminating instances before draining tasks have fully stopped.
Nomad Autoscaler v0.6.0 and v0.5.0 address both. New native APM plugins for Instana (v0.6.0) and InfluxDB (v0.5.0), alongside Prometheus instant query support (query_window = "instant"), let teams drive scaling decisions directly from application metrics. Schedule blocks let you define time-based windows to proactively warm capacity ahead of predictable events.
On lifecycle safety, v0.6.0 fixes a critical gap: the autoscaler now waits until all tasks on a drained node have fully stopped before issuing termination commands, preventing long-running shutdown hooks or background workers from being interrupted prematurely.
»Simplify GitOps and eliminate state secrets with Terraform provider v2.6.1
Managing Nomad jobs declaratively with Terraform is standard practice for modern platform teams. Two friction points have consistently come up: sensitive ACL tokens and variables persisting in Terraform state files, and Terraform applies overwriting dynamic task replica counts managed by the Nomad Autoscaler.
Terraform Provider v2.6.1 addresses both. Ephemeral Resources allow CI/CD pipelines to generate short-lived Nomad ACL tokens, client intro tokens, and variables on the fly, utilizing them during execution without storing sensitive secrets in permanent Terraform state. The preserve_counts parameter ensures that Terraform applies update job specifications, container images, and configurations without overwriting dynamic scaling counts established by the Nomad Autoscaler.
Combined with new nomad_node and nomad_nodes data sources for dynamic topology inspection, platform teams achieve clean, conflict-free GitOps automation.
»Nomad ecosystem releases at a glance
The following matrix summarizes the recent releases across the Nomad platform and ecosystem:
Component | Version | Key Operational & Architectural Value |
Nomad Platform | v2.1.1 / v2.0.5 | Batch Job Queue, raft-WAL store, plan_apply_pipeline, Snapshot Filtering |
Nomad Autoscaler | v0.6.0 / v0.5.0 | Instana & InfluxDB APM plugins, scheduled scaling windows, safe node termination |
Driver exec2 | v0.2.0 | Ambient Linux capabilities, CSI & Host volume mounts, alloc exec |
Driver Podman | v0.6.5 | Multi-arch overrides, HTTP(S) image transport, shared IPC, streamlined dependencies |
Driver Virt | v0.0.2-beta.1 | Native VM orchestration (Tech Preview) - Ceph/directory storage pools, block storage, macvtap |
Terraform Provider | v2.6.1 | Ephemeral secret resources, preserve_counts for seamless autoscaling, node data sources |
»Get started with Nomad 2.1
These releases reflect our ongoing focus: providing platform engineering teams with a reliable, flexible orchestrator that scales smoothly across any workload type.
Before upgrading, operators should review the Nomad 2.1.x Release Notes and Upgrade Guidance for detailed migration notes and plugin prerequisites.
Explore the documentation and downloads:
Download Nomad 2.1.1: releases.hashicorp.com/nomad/2.1.1
Explore Nomad Documentation: developer.hashicorp.com/nomad/docs
Join Community Discussions: https://discuss.hashicorp.com/c/nomad










