为MongoDB资源配置集群拓扑

您可以通过定义NodeAffinity和PodAffinity在Kubernetes中配置MongoDB资源的部署拓扑结构。 NodeAffinity和PodAffinity指定Kubernetes在Kubernetes集群中部署MongoDB自定义资源的位置。

You can implement node affinity by applying node labels to specific nodes within your Kubernetes cluster and defining labelSelector fields with corrosponding labels in your MongoDB CustomResourceDefinition. When you apply node labels and matching labelSelector values, Kubernetes only deploys a given MongoDB CustomResourceDefinition on the nodes you specify within your Kubernetes cluster. Implementing node affinity rules can be useful when deploying MongoDB resources in heterogeneous node groups, as it allows you to deploy specific resources on specific node varieties.

Similarly, you can implement Pod affinity by applying labels to Pods that are running in your cluster and aligning those label values with labelSelector values defined in your MongoDB CustomResourceDefinition. When you apply labels and matching labelSelector values, Kubernetes colocates MongoDB custom resource-managed Pods with Pods with matching labels applied. By collocating Pods, you can improve system performance and reduce latency between Pods that communicate regularly. You can also define Pod antiaffinity rules which allow you to specify Pods which should not be colocated.

For Standalone and Replica Set deployments, you can apply these affinity rules to the spec.podSpec of your MongoDB CustomResourceDefinition. For Sharded Cluster deployments, you can apply these affinity rules to the spec.configSrvPodSpec, spec.shardPodSpec, and spec.mongosPodSpec sections of your MongoDB CustomResourceDefinition.

您必须将MongoDB资源（例如分片集群部署中的 mongos、分片和配置服务器）部署在与MongoDB资源相同的命名空间中。但是，在该命名空间内，您可以在 ShardedCluster CustomResourceDefinition 中配置 mongos、分片和配置服务器资源类型的 nodeAffinity 和 podAffinity。

先决条件

要为MongoDB 部署配置部署拓扑结构，您必须通过MongoDB Kubernetes Operator 部署MongoDB副本集，并应用于Kubernetes资源的标签，这些标签将与MongoDB CustomResourceDefinition 中定义的 labelSelectors 保持一致，或者满足以下先决条件：

MongoDB Ops Manager实例或Cloud Manager组织。
Kubernetes Operator 的MongoDB控制器。
Kubernetes Operator ConfigMap。
Kubernetes Operator 的凭证，或配置其他密钥存储工具。
应用于Kubernetes资源的标签，这些标签将与MongoDB CustomResourceDefinition 中定义的 labelSelectors 保持一致。

步骤

更新您的MongoDB CRD 清单。

填充MongoDB分片集群定义的podTemplate.spec.affinity部分，如以下示例所示：

1 ---
2 apiVersion: mongodb.com/v1
3 kind: MongoDB
4 metadata:
5   name: my-sharded-cluster
6 spec:
7   shardCount: 2
8   mongodsPerShardCount: 3
9   mongosCount: 2
10   configServerCount: 3
11   version: 8.0.0
12   service: my-service
13 
14   opsManager:
15     configMapRef:
16       name: my-project
17   credentials: my-credentials
18   type: ShardedCluster
19 
20   persistent: true
21   configSrvPodSpec:
22     podTemplate:
23       spec:
24         affinity:
25           podAffinity:
26             requiredDuringSchedulingIgnoredDuringExecution:
27               - labelSelector:
28                   matchExpressions:
29                     - key: security
30                       operator: In
31                       values:
32                         - S1
33                 topologyKey: failure-domain.beta.kubernetes.io/zone
34           nodeAffinity:
35             requiredDuringSchedulingIgnoredDuringExecution:
36               nodeSelectorTerms:
37                 - matchExpressions:
38                     - key: kubernetes.io/e2e-az-name
39                       operator: In
40                       values:
41                         - e2e-az1
42                         - e2e-az2
43           podAntiAffinity:
44             requiredDuringSchedulingIgnoredDuringExecution:
45             - podAffinityTerm:
46                 topologyKey: nodeId
47   mongosPodSpec:
48     podTemplate:
49       spec:
50         affinity:
51           podAffinity:
52             requiredDuringSchedulingIgnoredDuringExecution:
53               - labelSelector:
54                   matchExpressions:
55                     - key: security
56                       operator: In
57                       values:
58                         - S1
59                 topologyKey: failure-domain.beta.kubernetes.io/zone
60           nodeAffinity:
61             requiredDuringSchedulingIgnoredDuringExecution:
62               nodeSelectorTerms:
63                 - matchExpressions:
64                     - key: kubernetes.io/e2e-az-name
65                       operator: In
66                       values:
67                         - e2e-az1
68                         - e2e-az2
69           podAntiAffinity:
70             requiredDuringSchedulingIgnoredDuringExecution:
71             - podAffinityTerm:
72                 topologyKey: nodeId
73   shardPodSpec:
74     podTemplate:
75       spec:
76         affinity:
77           podAffinity:
78             requiredDuringSchedulingIgnoredDuringExecution:
79               - labelSelector:
80                   matchExpressions:
81                     - key: security
82                       operator: In
83                       values:
84                         - S1
85                 topologyKey: failure-domain.beta.kubernetes.io/zone
86           nodeAffinity:
87             requiredDuringSchedulingIgnoredDuringExecution:
88               nodeSelectorTerms:
89                 - matchExpressions:
90                     - key: kubernetes.io/e2e-az-name
91                       operator: In
92                       values:
93                         - e2e-az1
94                         - e2e-az2
95           podAntiAffinity:
96             requiredDuringSchedulingIgnoredDuringExecution:
97             - podAffinityTerm:
98                 topologyKey: nodeId
99 ...

应用更改。

kubectl apply -f mongodb-crd.yaml

1	---
2	apiVersion: mongodb.com/v1
3	kind: MongoDB
4	metadata:
5	name: my-sharded-cluster
6	spec:
7	shardCount: 2
8	mongodsPerShardCount: 3
9	mongosCount: 2
10	configServerCount: 3
11	version: 8.0.0
12	service: my-service
13
14	opsManager:
15	configMapRef:
16	name: my-project
17	credentials: my-credentials
18	type: ShardedCluster
19
20	persistent: true
21	configSrvPodSpec:
22	podTemplate:
23	spec:
24	affinity:
25	podAffinity:
26	requiredDuringSchedulingIgnoredDuringExecution:
27	- labelSelector:
28	matchExpressions:
29	- key: security
30	operator: In
31	values:
32	- S1
33	topologyKey: failure-domain.beta.kubernetes.io/zone
34	nodeAffinity:
35	requiredDuringSchedulingIgnoredDuringExecution:
36	nodeSelectorTerms:
37	- matchExpressions:
38	- key: kubernetes.io/e2e-az-name
39	operator: In
40	values:
41	- e2e-az1
42	- e2e-az2
43	podAntiAffinity:
44	requiredDuringSchedulingIgnoredDuringExecution:
45	- podAffinityTerm:
46	topologyKey: nodeId
47	mongosPodSpec:
48	podTemplate:
49	spec:
50	affinity:
51	podAffinity:
52	requiredDuringSchedulingIgnoredDuringExecution:
53	- labelSelector:
54	matchExpressions:
55	- key: security
56	operator: In
57	values:
58	- S1
59	topologyKey: failure-domain.beta.kubernetes.io/zone
60	nodeAffinity:
61	requiredDuringSchedulingIgnoredDuringExecution:
62	nodeSelectorTerms:
63	- matchExpressions:
64	- key: kubernetes.io/e2e-az-name
65	operator: In
66	values:
67	- e2e-az1
68	- e2e-az2
69	podAntiAffinity:
70	requiredDuringSchedulingIgnoredDuringExecution:
71	- podAffinityTerm:
72	topologyKey: nodeId
73	shardPodSpec:
74	podTemplate:
75	spec:
76	affinity:
77	podAffinity:
78	requiredDuringSchedulingIgnoredDuringExecution:
79	- labelSelector:
80	matchExpressions:
81	- key: security
82	operator: In
83	values:
84	- S1
85	topologyKey: failure-domain.beta.kubernetes.io/zone
86	nodeAffinity:
87	requiredDuringSchedulingIgnoredDuringExecution:
88	nodeSelectorTerms:
89	- matchExpressions:
90	- key: kubernetes.io/e2e-az-name
91	operator: In
92	values:
93	- e2e-az1
94	- e2e-az2
95	podAntiAffinity:
96	requiredDuringSchedulingIgnoredDuringExecution:
97	- podAffinityTerm:
98	topologyKey: nodeId
99	...