Manage HugePages
FEATURE STATE: Kubernetes v1.8 alpha
This feature is currently in a alpha state, meaning:
- The version names contain alpha (e.g. v1alpha1).
- Might be buggy. Enabling the feature may expose bugs. Disabled by default.
- Support for feature may be dropped at any time without notice.
- The API may change in incompatible ways in a later software release without notice.
- Recommended for use only in short-lived testing clusters, due to increased risk of bugs and lack of long-term support.
Kubernetes supports the allocation and consumption of pre-allocated huge pages by applications in a Pod as an alpha feature. This page describes how users can consume huge pages and the current limitations.
- Before you begin
- API
- Huge page requests must equal the limits. This is the default if limits are
- NUMA locality guarnatees as a feature of quality of service.
Before you begin
- Kubernetes nodes must pre-allocate huge pages in order for the node to report its huge page capacity. A node may only pre-allocate huge pages for a single size.
- A special alpha feature gate
HugePageshas to be set to true across the system:--feature-gates=HugePages=true.
The nodes will automatically discover and report all huge page resources as a schedulable resource.
API
Huge pages can be consumed via container level resource requirements using the
resource name hugepages-<size>, where size is the most compact binary notation
using integer values supported on a particular node. For example, if a node
supports 2048KiB page sizes, it will expose a schedulable resource
hugepages-2Mi. Unlike CPU or memory, huge pages do not support overcommit.
apiVersion: v1
kind: Pod
metadata:
generateName: hugepages-volume-
spec:
containers:
- image: fedora:latest
command:
- sleep
- inf
name: example
volumeMounts:
- mountPath: /hugepages
name: hugepage
resources:
limits:
hugepages-2Mi: 100Mi
volumes:
- name: hugepage
emptyDir:
medium: HugePages
«««< HEAD
-
Huge page requests must equal the limits. This is the default if limits are
- Huge page requests must equal the limits. This is the default if limits are
master specified, but requests are not.
- Huge pages are isolated at a pod scope, container isolation is planned in a future iteration.
- EmptyDir volumes backed by huge pages may not consume more huge page memory than the pod request.
- Applications that consume huge pages via
shmget()withSHM_HUGETLBmust «««< HEAD run with a supplemental group that matchesproc/sys/vm/hugetlb_shm_group======= run with a supplemental group that matchesproc/sys/vm/hugetlb_shm_group.master
Future
- Support container isolation of huge pages in addition to pod isolation. «««< HEAD
-
NUMA locality guarnatees as a feature of quality of service.
- NUMA locality guarantees as a feature of quality of service.
master
- ResourceQuota support.
- LimitRange support.