hashicorp/nomad v1.11.2

on GitHub

SECURITY:

• build: Updated toolchain to Go 1.25.6 [GH-27439]
• build: Updated toolchain to Go 1.25.7 [GH-27468]

IMPROVEMENTS:

• acl: Add finer grain permissions for managing job submissions [GH-27287]
• build: Add dev-static and static-release build targets that disable CGO and offer statically-linked binaries [GH-27310]
• cli: Highlight missing driver message in alloc metrics output [GH-27416]
• cli: Improve command line completion of the sentinel apply command [GH-27335]
• cni: Added /usr/libexec/cni as an additional default path within the client.cni_path configuration option [GH-27336]
• cni: Search all paths in cni_path instead of stopping on first failure [GH-27336]
• deps: Migrate from archived dependency github.com/mitchellh/mapstructure to github.com/go-viper/mapstructure/v2 [GH-27444]
• docker: Added support for reserved-only memory oversubscription without a hard limit [GH-27354]
• exec: Added support for reserved-only memory oversubscription without a hard limit [GH-27354]
• fingerprint: Added support for reloading the cpu, memory, network, CNI plugin, and cloud provider fingerprints without restarting the client agent [GH-27452]
• qemu: adds an emulator allowlist to qemu plugin config [GH-27182]
• rpc: Submitting a plan no longer serializes the whole Job object [GH-27424]
• scheduler: Do not create node evals for terminal node allocs [GH-27423]
• scheduler: Do not create node evaluations for system jobs that are stopped [GH-27419]
• sentinel: Added a new nomad_var built-in import for fetching Nomad variables under the nomad/sentinel path for use in policy evaluation
• sentinel: Added opt-in support for the http module via the sentinel.additional_enabled_modules configuration
• state: avoid unneded allocation copy when building event payload [GH-27311]

BUG FIXES:

• acl: Fixed a bug where host-volume-delete capability was not allowed when writing a policy [GH-27434]
• api: exit EventStream.Stream on first error [GH-27141]
• api: only include running tasks in allocation resource usage [GH-27317]
• api: return proper 403 message when getting variables instead of swallowing error [GH-27269]
• artifact: Fixed a bug that prevented the sandbox from moving downloaded files to the target directory on Windows [GH-27398]
• checks: Fixed a bug where script checks with task-level interpolation would fail to heartbeat to Consul [GH-27453]
• client: Added a new fingerprint configuration block which allows users to specify retry behavior for the env_aws, env_azure, env_digitalocean and env_gcp fingerprinters. [GH-27161]
• client: Fix unchanged devices causing extraneous node updates [GH-27363]
• client: Fixed generation of the "NOMAD_ALLOC_ADDR_" environment variable when using static port assignments [GH-27305]
• core: Fixed a bug where follow-up evals could be created for failed evaluations of garbage collected jobs [GH-27367]
• deployment: Fixed a bug where deploying a system job could panic the leader [GH-27262]
• deployments: Fixed a bug where system deployments can violate update.max_parallel if another eval for the job is triggered while allocs are pending [GH-27284]
• disconnect: allocations with a disconnect.lost_after > 0 and replace = true will now follow the reschedule block instead of immediately being replaced. [GH-27053]
• dispatch: Fixed a bug where concurrent dispatch requests could ignore the idempotency token [GH-27353]
• drivers: adds hostname to NetworkCreateRequest for external drivers [GH-27273]
• event broker: fix memory leak in methods that close subscriptions [GH-27312]
• event stream: Fixed a bug where the HTTP handler can block forever and cause high memory usage if an API client reads too slowly from the stream [GH-27397]
• host volumes: Fixed a bug where allocations that request volumes with sticky=true could not be placed if previous allocations in the job claimed volumes [GH-27470]
• job: Correctly validate any constraint attributes to ensure they conform to known formats [GH-27355]
• keyring (Enterprise): Fixed a bug where servers configured with high availability keyrings with pre-1.9.0 keystores would not start if one of the external KMS was unreachable [GH-27279]
• multiregion: fixes a bug where resubmitting an unchanged job would cause server handler to hang [GH-27386]
• numa: Fixed a bug where NUMA detection would cause a panic on hosts with discontinuous node IDs [GH-27277]
• qemu: change driver filesystem isolation to "None" for proper variable interpolation in job spec [GH-27246]
• qemu: fixes graceful_shutdown to wait kill_timeout before signalling process [GH-27316]
• ui: Tagging job versions in another namespace than the default-namespace resulted in an error [GH-27282]
• ui: fix bug preventing OIDC login when iss parameter is required [GH-27248]