Nomad 2.1.0 Introduces Experimental Batch Job Queues for Node Pools

Nomad 2.1.0 introduces an experimental batch job queue for node pools, configurable per node pool with FIFO and Dynamic Priority options.

Upgrading requires attention to breaking changes. Joining a server through the CLI or API now needs the agent:write permission. Additionally, the deprecated server-level retry_join, retry_max, retry_interval, and start_join blocks are removed, meaning you must move those fields into the server_join block.

API changes include the deprecation of Resources and TaskResources fields on the Allocation struct in favor of AllocatedResources. Client state databases restored from backups older than Nomad 0.9 are no longer compatible and cause client startup to fail.

The CLI adds an operator snapshot filter command. Jobspec2 supports optional() object type attributes in variables, and TLS configuration cleanup may alter TLS handshake failure error messages.

Fixes address node heartbeats sent from other nodes, optimistic snapshot refreshing in the plan applier under sustained saturation, Apple M4 CPU client fingerprinting, a task state overwrite bug on clients, and a CLI panic on invalid inputs for var put. The CSI plugin status command no longer lists unsupported capabilities with -verbose, and UI graph and paging bugs are resolved.

Builds update to Go 1.27.2 and 1.27.3, and the container image busybox moves to 1.38.0.