commit
62f26c69e2
|
@ -18,10 +18,9 @@ content_type: task
|
|||
|
||||
<!--
|
||||
Kubernetes supports the allocation and consumption of pre-allocated huge pages
|
||||
by applications in a Pod as a **GA** feature. This page describes how users
|
||||
can consume huge pages and the current limitations.
|
||||
by applications in a Pod. This page describes how users can consume huge pages.
|
||||
--->
|
||||
作为 **GA** 特性,Kubernetes 支持在 Pod 应用中使用预先分配的巨页。本文描述了用户如何使用巨页,以及当前的限制。
|
||||
Kubernetes 支持在 Pod 应用中使用预先分配的巨页。本文描述了用户如何使用巨页,以及当前的限制。
|
||||
|
||||
|
||||
|
||||
|
@ -48,29 +47,70 @@ schedulable resource.
|
|||
|
||||
<!--
|
||||
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. Note
|
||||
that when requesting hugepage resources, either memory or CPU resources must
|
||||
be requested as well.
|
||||
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 and 1048576KiB page sizes, it will expose a schedulable
|
||||
resources `hugepages-2Mi` and `hugepages-1Gi`. Unlike CPU or memory, huge pages
|
||||
do not support overcommit. Note that when requesting hugepage resources, either
|
||||
memory or CPU resources must be requested as well.
|
||||
|
||||
A pod may consume multiple huge page sizes in a single pod spec. In this case it
|
||||
must use `medium: HugePages-<hugepagesize>` notation for all volume mounts.
|
||||
--->
|
||||
|
||||
用户可以通过在容器级别的资源需求中使用资源名称 `hugepages-<size>` 来使用巨页,其中的 size 是特定节点上支持的以整数值表示的最小二进制单位。 例如,如果节点支持 2048KiB 的页面规格, 它将暴露可供调度的资源 `hugepages-2Mi`。 与 CPU 或内存不同,巨页不支持过量使用(overcommit)。
|
||||
注意,在请求巨页资源时,还必须请求内存或 CPU 资源。
|
||||
用户可以通过在容器级别的资源需求中使用资源名称 `hugepages-<size>` 来使用巨页,其中的 size 是特定节点上支持的以整数值表示的最小二进制单位。 例如,如果一个节点支持 2048KiB 和 1048576KiB 页面大小,它将公开可调度的资源 `hugepages-2Mi` 和 `hugepages-1Gi`。与 CPU 或内存不同,巨页不支持过量使用(overcommit)。注意,在请求巨页资源时,还必须请求内存或 CPU 资源。
|
||||
|
||||
同一 Pod 的 spec 中可能会消耗不同尺寸的巨页。在这种情况下,它必须对所有挂载卷使用 `medium: HugePages-<hugepagesize>` 标识。
|
||||
|
||||
```yaml
|
||||
apiVersion: v1
|
||||
kind: Pod
|
||||
metadata:
|
||||
generateName: hugepages-volume-
|
||||
name: huge-pages-example
|
||||
spec:
|
||||
containers:
|
||||
- image: fedora:latest
|
||||
- name: example
|
||||
image: fedora:latest
|
||||
command:
|
||||
- sleep
|
||||
- inf
|
||||
volumeMounts:
|
||||
- mountPath: /hugepages-2Mi
|
||||
name: hugepage-2mi
|
||||
- mountPath: /hugepages-1Gi
|
||||
name: hugepage-1gi
|
||||
resources:
|
||||
limits:
|
||||
hugepages-2Mi: 100Mi
|
||||
hugepages-1Gi: 2Gi
|
||||
memory: 100Mi
|
||||
requests:
|
||||
memory: 100Mi
|
||||
volumes:
|
||||
- name: hugepage-2mi
|
||||
emptyDir:
|
||||
medium: HugePages-2Mi
|
||||
- name: hugepage-1gi
|
||||
emptyDir:
|
||||
medium: HugePages-1Gi
|
||||
```
|
||||
<!--
|
||||
A pod may use `medium: HugePages` only if it requests huge pages of one size.
|
||||
-->
|
||||
Pod 只有在请求同一大小的巨页时才使用 `medium:HugePages`。
|
||||
|
||||
```yaml
|
||||
apiVersion: v1
|
||||
kind: Pod
|
||||
metadata:
|
||||
name: huge-pages-example
|
||||
spec:
|
||||
containers:
|
||||
- name: example
|
||||
image: fedora:latest
|
||||
command:
|
||||
- sleep
|
||||
- inf
|
||||
name: example
|
||||
volumeMounts:
|
||||
- mountPath: /hugepages
|
||||
name: hugepage
|
||||
|
@ -89,8 +129,7 @@ spec:
|
|||
<!--
|
||||
- Huge page requests must equal the limits. This is the default if limits are
|
||||
specified, but requests are not.
|
||||
- Huge pages are isolated at a pod scope, container isolation is planned in a
|
||||
future iteration.
|
||||
- Huge pages are isolated at a container scope, so each container has own limit on their cgroup sandbox as requested in a container spec.
|
||||
- EmptyDir volumes backed by huge pages may not consume more huge page memory
|
||||
than the pod request.
|
||||
- Applications that consume huge pages via `shmget()` with `SHM_HUGETLB` must
|
||||
|
@ -98,30 +137,25 @@ spec:
|
|||
- Huge page usage in a namespace is controllable via ResourceQuota similar
|
||||
to other compute resources like `cpu` or `memory` using the `hugepages-<size>`
|
||||
token.
|
||||
- Support of multiple sizes huge pages is feature gated. It can be
|
||||
enabled with the `HugePageStorageMediumSize` [feature
|
||||
gate](/docs/reference/command-line-tools-reference/feature-gates/) on the {{<
|
||||
glossary_tooltip text="kubelet" term_id="kubelet" >}} and {{<
|
||||
glossary_tooltip text="kube-apiserver"
|
||||
term_id="kube-apiserver" >}} (`--feature-gates=HugePageStorageMediumSize=true`).
|
||||
--->
|
||||
|
||||
- 巨页的资源请求值必须等于其限制值。该条件在指定了资源限制,而没有指定请求的情况下默认成立。
|
||||
- 巨页是被隔离在 pod 作用域的,计划在将来的迭代中实现容器级别的隔离。
|
||||
- 巨页是被隔离在 pod 作用域的,因此每个容器在 spec 中都对 cgroup 沙盒有自己的限制。
|
||||
- 巨页可用于 EmptyDir 卷,不过 EmptyDir 卷所使用的巨页数量不能够超出 Pod 请求的巨页数量。
|
||||
- 通过带有 `SHM_HUGETLB` 的 `shmget()` 使用巨页的应用,必须运行在一个与
|
||||
`proc/sys/vm/hugetlb_shm_group` 匹配的补充组下。
|
||||
- 通过 ResourceQuota 资源,可以使用 `hugepages-<size>` 标记控制每个命名空间下的巨页使用量,
|
||||
类似于使用 `cpu` 或 `memory` 来控制其他计算资源。
|
||||
|
||||
<!--
|
||||
## Future
|
||||
|
||||
- Support container isolation of huge pages in addition to pod isolation.
|
||||
- NUMA locality guarantees as a feature of quality of service.
|
||||
- LimitRange support.
|
||||
--->
|
||||
|
||||
## 待实现的特性
|
||||
|
||||
- 在 pod 级别隔离的基础上,支持巨页在容器级别的隔离。
|
||||
- 作为服务质量特性,保证巨页的 NUMA 局部性。
|
||||
- 支持 LimitRange 。
|
||||
|
||||
- 多种尺寸的巨页的支持需要特性门控配置。它可以通过 `HugePageStorageMediumSize` [特性门控](/zh/docs/reference/command-line-tools-reference/feature-gates/)在 {{<
|
||||
glossary_tooltip text="kubelet" term_id="kubelet" >}} 和 {{<
|
||||
glossary_tooltip text="kube-apiserver"
|
||||
term_id="kube-apiserver" >}} 中开启(`--feature-gates=HugePageStorageMediumSize=true`)。
|
||||
|
||||
|
||||
|
||||
|
|
Loading…
Reference in New Issue