Skip to content

Commit d29693f

Browse files
k8s-ci-robotk8s-ci-robot
authored
Merge pull request #27800 from zhiguo-lu/zh-trans-concepts-topology-aware-hints
[zh] translate concepts/Topology Aware Hints
2 parents d52a4a3 + 94e8829 commit d29693f

File tree

1 file changed

+285
-0
lines changed

1 file changed

+285
-0
lines changed

‎content/zh/docs/concepts/services-networking/topology-aware-hints.md

Lines changed: 285 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,285 @@
1+
---
2+
title: 拓扑感知提示
3+
content_type: concept
4+
weight: 45
5+
---
6+
<!--
7+
---
8+
reviewers:
9+
- robscott
10+
title: Topology Aware Hints
11+
content_type: concept
12+
weight: 45
13+
---
14+
-->
15+
16+
<!-- overview -->
17+
18+
{{< feature-state for_k8s_version="v1.21" state="alpha" >}}
19+
20+
<!--
21+
_Topology Aware Hints_ enable topology aware routing by including suggestions
22+
for how clients should consume endpoints. This approach adds metadata to enable
23+
consumers of EndpointSlice and / or and Endpoints objects, so that traffic to
24+
those network endpoints can be routed closer to where it originated.
25+
26+
For example, you can route traffic within a locality to reduce
27+
costs, or to improve network performance.
28+
-->
29+
_拓扑感知提示_ 包含客户怎么使用服务端点的建议，从而实现了拓扑感知的路由功能。
30+
这种方法添加了元数据，以启用 EndpointSlice 和/或 Endpoints 对象的调用者，
31+
这样，访问这些网络端点的请求流量就可以在它的发起点附近就近路由。
32+
33+
例如，你可以在一个地域内路由流量，以降低通信成本，或提高网络性能。
34+
35+
<!-- body -->
36+
37+
<!--
38+
## Motivation
39+
-->
40+
## 动机 {#motivation}
41+
42+
<!--
43+
Kubernetes clusters are increasingly deployed in multi-zone environments.
44+
_Topology Aware Hints_ provides a mechanism to help keep traffic within the zone
45+
it originated from. This concept is commonly referred to as "Topology Aware
46+
Routing". When calculating the endpoints for a {{< glossary_tooltip term_id="Service" >}},
47+
the EndpointSlice controller considers the topology (region and zone) of each endpoint
48+
and populates the hints field to allocate it to a zone.
49+
Cluster components such as the {{< glossary_tooltip term_id="kube-proxy" text="kube-proxy" >}}
50+
can then consume those hints, and use them to influence how traffic to is routed
51+
(favoring topologically closer endpoints).
52+
-->
53+
Kubernetes 集群越来越多的部署到多区域环境中。
54+
_拓扑感知提示_ 提供了一种把流量限制在它的发起区域之内的机制。
55+
这个概念一般被称之为 “拓扑感知路由”。
56+
在计算 {{< glossary_tooltip term_id="Service" >}} 的端点时，
57+
EndpointSlice 控制器会评估每一个端点的拓扑（地域和区域），填充提示字段，并将其分配到某个区域。
58+
集群组件，例如{{< glossary_tooltip term_id="kube-proxy" text="kube-proxy" >}}
59+
就可以使用这些提示信息，并用他们来影响流量的路由（倾向于拓扑上相邻的端点）。
60+
61+
<!--
62+
## Using Topology Aware Hints
63+
-->
64+
## 使用拓扑感知提示 {#using-topology-aware-hints}
65+
66+
<!--
67+
If you have [enabled](/docs/tasks/administer-cluster/enabling-topology-aware-hints) the
68+
overall feature, you can activate Topology Aware Hints for a Service by setting the
69+
`service.kubernetes.io/topology-aware-hints` annotation to `auto`. This tells
70+
the EndpointSlice controller to set topology hints if it is deemed safe.
71+
Importantly, this does not guarantee that hints will always be set.
72+
-->
73+
如果你已经[启用](/zh/docs/tasks/administer-cluster/enabling-topology-aware-hints)了整个特性，
74+
就可以通过把注解 `service.kubernetes.io/topology-aware-hints` 的值设置为 `auto`，
75+
来激活服务的拓扑感知提示功能。
76+
这告诉 EndpointSlice 控制器在它认为安全的时候来设置拓扑提示。
77+
重要的是，这并不能保证总会设置提示（hints）。
78+
79+
<!--
80+
## How it works {#implementation}
81+
-->
82+
## 工作原理 {#implementation}
83+
84+
<!--
85+
The functionality enabling this feature is split into two components: The
86+
EndpointSlice controller and the kube-proxy. This section provides a high level overview
87+
of how each component implements this feature.
88+
-->
89+
此特性启用的功能分为两个组件：EndpointSlice 控制器和 kube-proxy。
90+
本节概述每个组件如何实现此特性。
91+
92+
<!--
93+
### EndpointSlice controller {#implementation-control-plane}
94+
-->
95+
### EndpointSlice 控制器 {#implementation-control-plane}
96+
97+
<!--
98+
The EndpointSlice controller is responsible for setting hints on EndpointSlices
99+
when this feature is enabled. The controller allocates a proportional amount of
100+
endpoints to each zone. This proportion is based on the
101+
[allocatable](/docs/tasks/administer-cluster/reserve-compute-resources/#node-allocatable)
102+
CPU cores for nodes running in that zone. For example, if one zone had 2 CPU
103+
cores and another zone only had 1 CPU core, the controller would allocated twice
104+
as many endpoints to the zone with 2 CPU cores.
105+
-->
106+
此特性开启后，EndpointSlice 控制器负责在 EndpointSlice 上设置提示信息。
107+
控制器按比例给每个区域分配一定比例数量的端点。
108+
这个比例来源于此区域中运行节点的
109+
[可分配](/zh/docs/tasks/administer-cluster/reserve-compute-resources/#node-allocatable)
110+
CPU 核心数。
111+
例如，如果一个区域拥有 2 CPU 核心，而另一个区域只有 1 CPU 核心，
112+
那控制器将给那个有 2 CPU 的区域分配两倍数量的端点。
113+
114+
<!--
115+
The following example shows what an EndpointSlice looks like when hints have
116+
been populated:
117+
-->
118+
以下示例展示了提供提示信息后 EndpointSlice 的样子：
119+
120+
```yaml
121+
apiVersion: discovery.k8s.io/v1
122+
kind: EndpointSlice
123+
metadata:
124+
name: example-hints
125+
labels:
126+
kubernetes.io/service-name: example-svc
127+
addressType: IPv4
128+
ports:
129+
- name: http
130+
protocol: TCP
131+
port: 80
132+
endpoints:
133+
- addresses:
134+
- "10.1.2.3"
135+
conditions:
136+
ready: true
137+
hostname: pod-1
138+
zone: zone-a
139+
hints:
140+
forZones:
141+
- name: "zone-a"
142+
```
143+
144+
### kube-proxy {#implementation-kube-proxy}
145+
146+
<!--
147+
The kube-proxy component filters the endpoints it routes to based on the hints set by
148+
the EndpointSlice controller. In most cases, this means that the kube-proxy is able
149+
to route traffic to endpoints in the same zone. Sometimes the controller allocates endpoints
150+
from a different zone to ensure more even distribution of endpoints between zones.
151+
This would result in some traffic being routed to other zones.
152+
-->
153+
kube-proxy 组件依据 EndpointSlice 控制器设置的提示，过滤由它负责路由的端点。
154+
在大多数场合，这意味着 kube-proxy 可以把流量路由到同一个区域的端点。
155+
有时，控制器从某个不同的区域分配端点，以确保在多个区域之间更平均的分配端点。
156+
这会导致部分流量被路由到其他区域。
157+
158+
<!--
159+
## Safeguards
160+
-->
161+
## 保护措施 {#safeguards}
162+
163+
<!--
164+
The Kubernetes control plane and the kube-proxy on each node apply some
165+
safeguard rules before using Topology Aware Hints. If these don't check out,
166+
the kube-proxy selects endpoints from anywhere in your cluster, regardless of the
167+
zone.
168+
-->
169+
Kubernetes 控制平面和每个节点上的 kube-proxy，在使用拓扑感知提示功能前，会应用一些保护措施规则。
170+
如果没有检出，kube-proxy 将无视区域限制，从集群中的任意节点上选择端点。
171+
172+
<!--
173+
1. **Insufficient number of endpoints:** If there are less endpoints than zones
174+
in a cluster, the controller will not assign any hints.
175+
-->
176+
1. **端点数量不足：** 如果一个集群中，端点数量少于区域数量，控制器不创建任何提示。
177+
178+
<!--
179+
2. **Impossible to achieve balanced allocation:** In some cases, it will be
180+
impossible to achieve a balanced allocation of endpoints among zones. For
181+
example, if zone-a is twice as large as zone-b, but there are only 2
182+
endpoints, an endpoint allocated to zone-a may receive twice as much traffic
183+
as zone-b. The controller does not assign hints if it can't get this "expected
184+
overload" value below an acceptable threshold for each zone. Importantly this
185+
is not based on real-time feedback. It is still possible for individual
186+
endpoints to become overloaded.
187+
-->
188+
2. **不可能实现均衡分配：** 在一些场合中，不可能实现端点在区域中的平衡分配。
189+
例如，假设 zone-a 比 zone-b 大两倍，但只有 2 个端点，
190+
那分配到 zone-a 的端点可能收到比 zone-b多两倍的流量。
191+
如果控制器不能确定此“期望的过载”值低于每一个区域可接受的阈值，控制器将不指派提示信息。
192+
重要的是，这不是基于实时反馈。所以对于单独的端点仍有可能超载。
193+
194+
<!--
195+
3. **One or more Nodes has insufficient information:** If any node does not have
196+
a `topology.kubernetes.io/zone` label or is not reporting a value for
197+
allocatable CPU, the control plane does not set any topology-aware endpoint
198+
hints and so kube-proxy does not filter endpoints by zone.
199+
-->
200+
3. **一个或多个节点信息不足：** 如果任一节点没有设置标签 `topology.kubernetes.io/zone`，
201+
或没有上报可分配的 CPU 数据，控制平面将不会设置任何拓扑感知提示，
202+
继而 kube-proxy 也就不能通过区域过滤端点。
203+
204+
<!--
205+
4. **One or more endpoints does not have a zone hint:** When this happens,
206+
the kube-proxy assumes that a transition from or to Topology Aware Hints is
207+
underway. Filtering endpoints for a Service in this state would be dangerous
208+
so the kube-proxy falls back to using all endpoints.
209+
-->
210+
4. **一个或多个端点没有设置区域提示：** 当这类事情发生时，
211+
kube-proxy 会假设这是正在执行一个从/到拓扑感知提示的转移。
212+
在这种场合下过滤Service 的端点是有风险的，所以 kube-proxy 回撤为使用所有的端点。
213+
214+
<!--
215+
5. **A zone is not represented in hints:** If the kube-proxy is unable to find
216+
at least one endpoint with a hint targeting the zone it is running in, it falls
217+
to using endpoints from all zones. This is most likely to happen as you add
218+
a new zone into your existing cluster.
219+
-->
220+
5. **不在提示中的区域：** 如果 kube-proxy 不能根据一个指示在它所在的区域中发现一个端点，
221+
它回撤为使用所有节点的端点。当你的集群新增一个新的区域时，这种情况发生概率很高。
222+
223+
<!--
224+
## Constraints
225+
-->
226+
## 限制 {#constraints}
227+
228+
<!--
229+
* Topology Aware Hints are not used when either `externalTrafficPolicy` or
230+
`internalTrafficPolicy` is set to `Local` on a Service. It is possible to use
231+
both features in the same cluster on different Services, just not on the same
232+
Service.
233+
-->
234+
* 当 Service 的 `externalTrafficPolicy` 或 `internalTrafficPolicy` 设置值为 `Local` 时，
235+
拓扑感知提示功能不可用。
236+
你可以在一个集群的不同服务中使用这两个特性，但不能在同一个服务中这么做。
237+
238+
<!--
239+
* This approach will not work well for Services that have a large proportion of
240+
traffic originating from a subset of zones. Instead this assumes that incoming
241+
traffic will be roughly proportional to the capacity of the Nodes in each
242+
zone.
243+
-->
244+
* 这种方法不适用于大部分流量来自于一部分区域的服务。
245+
相反的，这里假设入站流量将根据每个区域中节点的服务能力按比例的分配。
246+
247+
<!--
248+
* The EndpointSlice controller ignores unready nodes as it calculates the
249+
proportions of each zone. This could have unintended consequences if a large
250+
portion of nodes are unready.
251+
-->
252+
* EndpointSlice 控制器在计算每一个区域的容量比例时，会忽略未就绪的节点。
253+
在大量节点未就绪的场景下，这样做会带来非预期的结果。
254+
255+
<!--
256+
* The EndpointSlice controller does not take into account {{< glossary_tooltip
257+
text="tolerations" term_id="toleration" >}} when deploying calculating the
258+
proportions of each zone. If the Pods backing a Service are limited to a
259+
subset of Nodes in the cluster, this will not be taken into account.
260+
-->
261+
* EndpointSlice 控制器在计算每一个区域的部署比例时，并不会考虑
262+
{{< glossary_tooltip text="容忍度" term_id="toleration" >}}。
263+
如果服务后台的 Pod 被限制只能运行在集群节点的一个子集上，这些信息并不会被使用。
264+
265+
<!--
266+
* This may not work well with autoscaling. For example, if a lot of traffic is
267+
originating from a single zone, only the endpoints allocated to that zone will
268+
be handling that traffic. That could result in {{< glossary_tooltip
269+
text="Horizontal Pod Autoscaler" term_id="horizontal-pod-autoscaler" >}}
270+
either not picking up on this event, or newly added pods starting in a
271+
different zone.
272+
-->
273+
* 这种方法和自动扩展机制之间不能很好的协同工作。例如，如果大量流量来源于一个区域，
274+
那只有分配到该区域的端点才可用来处理流量。这会导致
275+
{{< glossary_tooltip text="Pod 自动水平扩展" term_id="horizontal-pod-autoscaler" >}}
276+
要么不能拾取此事件，要么新增 Pod 被启动到其他区域。
277+
278+
## {{% heading "whatsnext" %}}
279+
280+
<!--
281+
* Read about [enabling Topology Aware Hints](/docs/tasks/administer-cluster/enabling-topology-aware-hints/)
282+
* Read [Connecting Applications with Services](/docs/concepts/services-networking/connect-applications-service/)
283+
-->
284+
* 参阅[启用拓扑感知提示](/zh/docs/tasks/administer-cluster/enabling-topology-aware-hints/)
285+
* 参阅[通过服务连通应用](/zh/docs/concepts/services-networking/connect-applications-service/)

0 commit comments

Comments
 (0)