apache · funky-eyes · Oct 28, 2024 · Oct 13, 2024 · Oct 14, 2024 · Oct 19, 2024
diff --git a/blog/seata-namingserver.md b/blog/seata-namingserver.md
@@ -0,0 +1 @@
+Placeholder. DO NOT DELETE.
diff --git a/docs/user/registry/namingserver.md b/docs/user/registry/namingserver.md
@@ -0,0 +1 @@
+Placeholder. DO NOT DELETE.
diff --git a/i18n/en/docusaurus-plugin-content-docs/current/user/registry/namingserver.md b/i18n/en/docusaurus-plugin-content-docs/current/user/registry/namingserver.md
@@ -0,0 +1,82 @@
+---
+Title: Seata Namingserver Registration Center
+Keywords: [ Seata, Namingserver ]
+Description: Namingserver registration center.
+---
+
+
+# Namingserver Registration Center
+Namingserver is the native registration center of Seata.
+
+## Preparations
+You can choose to run the naming server in the compiler or after packaging it.
+
+### Running the naming server in the compiler
+Go to the namingserver directory, set the port number for the naming server in resources/application.yml, and start the naming server.
+
+### Running Naming Server
+Download the Seata 2.2.0 binary package from the [link](https://seata.apache.org/unversioned/download/seata-server/) and unzip it. Then, enter the seata-namingserver directory.
+
+In the conf/application.yml file, configure the port number for starting the naming server.
+
+For Mac or Linux, run:
+
+```shell
+bin/seata-namingserver.sh
+```
+For Windows, run:
+
+```shell
+bin/seata-namingserver.bat
+```
+
+## Getting Started Quickly
+The process to use the naming server as Seata's registration center is very simple and involves configuring both the client and server sides.
+
+### Client-side Configuration for the Registration Center
+Add the corresponding registration center in the application.yml:
+
+```yaml
+seata:
+  registry:
+    type: seata
+  seata:
+    server-addr: 127.0.0.1:8081   ## IP + port of the naming server configured in the first step. Use commas to separate multiple naming server nodes.
+    namespace: public  ## Namespace
+    heartbeat-period: 5000  ## Heartbeat interval
+```
+### Server-side Configuration for the Registration Center
+Add the following configuration to `conf/application.yaml`, with the rest of the configuration referring to configuration options:
+
+```yaml
+seata:
+  registry:
+    type: seata
+  seata:
+    server-addr: 127.0.0.1:8081   ## IP + port of the naming server configured in the first step. Use commas to separate multiple naming server nodes.
+    cluster: default  ## Cluster name
+    namespace: public  ## Namespace
+    heartbeat-period: 5000  ## Heartbeat interval
+```
+### Create a Mapping Between the Client's Transaction Group and the Seata Cluster
+Send an HTTP request to one naming server node to create a transaction group-to-cluster mapping (the naming server node will automatically synchronize with other nodes):
+
+```shell
+http://127.0.0.1:8081/naming/v1/addGroup?clusterName=cluster2&namespace=public&unitName&vGroup=my_test_tx_group
+```
+(In this, namespace is the namespace configured on the client side, vGroup is the transaction group configured on the client side, and clusterName is the cluster name on the server side that needs to be mapped.)
+
+
+### Change the Mapping Between the Client's Transaction Group and the Seata Cluster
+Send an HTTP request to one naming server node to change the transaction group-to-cluster mapping (the naming server node will automatically synchronize with other nodes):
+
+```shell
+http://127.0.0.1:8081/naming/v1/changeGroup?clusterName=cluster2&namespace=public&unitName&vGroup=my_test_tx_group
+```
+(In this, namespace is the namespace configured on the client side, vGroup is the transaction group configured on the client side, and clusterName is the cluster name on the server side that needs to be mapped.)
+
+Afterward, once Seata-Server is started and the client configuration is complete, you can begin to experience Seata services.
+
+Tips:
+- 1.Please ensure that the client and server are registered under the same namespace; otherwise, the service will not be found.
+- 2.Note that the naming server is only allowed to be used in a private network; do not expose it to the public environment.
diff --git a/i18n/en/docusaurus-plugin-content-docs/version-v2.2/user/registry/namingserver.md b/i18n/en/docusaurus-plugin-content-docs/version-v2.2/user/registry/namingserver.md
@@ -0,0 +1,84 @@
+---
+Title: Seata Namingserver Registration Center
+Keywords: [ Seata, Namingserver ]
+Description: Namingserver registration center.
+---
+
+
+# Namingserver Registration Center
+Namingserver is the native registration center of Seata.
+
+## Preparations
+You can choose to run the naming server in the compiler or after packaging it.
+
+### Running the naming server in the compiler
+Go to the namingserver directory, set the port number for the naming server in resources/application.yml, and start the naming server.
+
+### Running the naming server in the compiler
+Go to the namingserver directory, set the port number for the naming server in resources/application.yml, and start the naming server.
+
+### Running Naming Server
+Download the Seata 2.2.0 binary package from the [link](https://seata.apache.org/unversioned/download/seata-server/) and unzip it. Then, enter the seata-namingserver directory.
+
+In the conf/application.yml file, configure the port number for starting the naming server.
+
+For Mac or Linux, run:
+
+```shell
+bin/seata-namingserver.sh
+```
+For Windows, run:
+
+```shell
+bin/seata-namingserver.bat
+```
+## Getting Started Quickly
+The process to use the naming server as Seata's registration center is very simple and involves configuring both the client and server sides.
+
+### Client-side Configuration for the Registration Center
+Add the corresponding registration center in the application.yml:
+
+```yaml
+seata:
+  registry:
+    type: seata
+  seata:
+    server-addr: 127.0.0.1:8081   ## IP + port of the naming server configured in the first step. Use commas to separate multiple naming server nodes.
+    namespace: public  ## Namespace
+    heartbeat-period: 5000  ## Heartbeat interval
+```
+### Server-side Configuration for the Registration Center
+Add the following configuration to `conf/application.yaml`, with the rest of the configuration referring to configuration options:
+
+```yaml
+seata:
+  registry:
+    type: seata
+  seata:
+    server-addr: 127.0.0.1:8081   ## IP + port of the naming server configured in the first step. Use commas to separate multiple naming server nodes.
+    cluster: default  ## Cluster name
+    namespace: public  ## Namespace
+    heartbeat-period: 5000  ## Heartbeat interval
+```
+### Create a Mapping Between the Client's Transaction Group and the Seata Cluster
+Send an HTTP request to one naming server node to create a transaction group-to-cluster mapping (the naming server node will automatically synchronize with other nodes):
+
+```shell
+http://127.0.0.1:8081/naming/v1/addGroup?clusterName=cluster2&namespace=public&unitName&vGroup=my_test_tx_group
+```
+(In this, namespace is the namespace configured on the client side, vGroup is the transaction group configured on the client side, and clusterName is the cluster name on the server side that needs to be mapped.)
+
+
+### Change the Mapping Between the Client's Transaction Group and the Seata Cluster
+Send an HTTP request to one naming server node to change the transaction group-to-cluster mapping (the naming server node will automatically synchronize with other nodes):
+
+```shell
+http://127.0.0.1:8081/naming/v1/changeGroup?clusterName=cluster2&namespace=public&unitName&vGroup=my_test_tx_group
+```
+(In this, namespace is the namespace configured on the client side, vGroup is the transaction group configured on the client side, and clusterName is the cluster name on the server side that needs to be mapped.)
+
+Afterward, once Seata-Server is started and the client configuration is complete, you can begin to experience Seata services.
+
+Tips:
+- 1.Please ensure that the client and server are registered under the same namespace; otherwise, the service will not be found.
+- 2.Note that the naming server is only allowed to be used in a private network; do not expose it to the public environment.
diff --git a/i18n/zh-cn/docusaurus-plugin-content-blog/seata-namingserver.md b/i18n/zh-cn/docusaurus-plugin-content-blog/seata-namingserver.md
@@ -0,0 +1,155 @@
+---
+title: Seata Namingserver
+author: 蒋俊敏
+description: 在本文中，我将分享seata的注册中心namingserver的设计思路以及使用方式
+date: 2024/9/25
+keywords: [seata,分布式事务,注册中心,namingserver]
+---
+
+# 1. 项目背景
+seata目前支持多种注册中心的实现，为了提供整个链路的闭环功能，seata设计推出了原生的注册中心namingserver
+
+# 2. 领域模型
+
+### 2.1 命名空间与事务分组
+
+- Namespace：在NamingServer模型中，命名空间（Namespace）用于实现命名空间的环境隔离。它允许在不同的环境（如开发、测试、生产）中隔离各自的服务实例。
+- Cluster与Unit：Cluster（集群）负责事务分组的处理，Unit则在每个集群内部做负载均衡。事务分组（vgroup）在命名空间和集群的配合下，通过元数据定位到具体的TC节点。
+
+### 2.2 事务处理流程与namingserver的交互
+
+![img](/img/blog/namingserver1.png)
+事务处理流程与namingserver的交互的流程如下：
+
+1.在client侧配置好Namingserver的地址和相关配置
+
+2.client启动后TM向namingserver发起服务发现的请求
+
+3.namingserver根据TM传来的vGroup参数和内存中的事务分组映射关系返回相关的集群列表，namingserver返回的集群列表元数据如下
+
+```json
+{
+    "clusterList": [
+        {
+            "clusterName": "cluster2",
+            "clusterType": "default",
+          	"groupList":[group1,group2]
+            "unitData": [
+                {
+                    "unitName": "115482ee-cf27-45d6-b17e-31b9e2d7892f",
+                    "namingInstanceList": [
+                        {
+                            "ip": "172.31.31.191",
+                            "port": 8092,
+                            "nettyPort": 0,
+                            "grpcPort": 0,
+                            "weight": 1.0,
+                            "healthy": true,
+                            "timeStamp": 1695042063334,
+                            "role": member,
+                            "metadata": {
+                                "weight": 1,
+                                "cluster-type": "default"
+                            }
+                        }
+                    ]
+                },
+                {
+                    "unitName": "097e6ab7-d2d2-47e4-a578-fae1a4f4c517",
+                    "namingInstanceList": [
+                        {
+                            "ip": "172.31.31.191",
+                            "port": 8091,
+                            "nettyPort": 0,
+                            "grpcPort": 0,
+                            "weight": 1.0,
+                            "healthy": true,
+                            "timeStamp": 1695042076481,
+                            "role": member,
+                            "metadata": {
+                                "weight": 1,
+                                "cluster-type": "default"
+                            }
+                        }
+                    ]
+                }
+            ]
+        }
+    ],
+    "term": 1695042076578
+}
+```
+
+4.客户端通过负载均衡策略找出合适的TC节点开启事务
+
+5.TM将事务分组和TC节点传递给RM
+
+6.RM向TC节点发起分支注册的请求
+
+7.TC节点完成二阶段下发
+
+# 3.设计思路
+
+### 3.1 AP还是CP？
+
+CAP协议又称CAP定理，指的是在一个分布式系统中，Consistency（一致性）、 Availability（可用性）、Partition tolerance（分区容错性），三者不可得兼。 分布式系统的CAP理论：理论首先把分布式系统中的三个特性进行了如下归纳：
+
+● 一致性（C）：在分布式系统中的所有数据备份，在同一时刻是否同样的值。（等同于所有节点访问同一份最新的数据副本）
+
+● 可用性（A）：在集群中一部分节点故障后，集群整体是否还能响应客户端的读写请求。（对数据更新具备高可用性）
+
+● 分区容错性（P）：以实际效果而言，分区相当于对通信的时限要求。系统如果不能在时限内达成数据一致性，就意味着发生了分区的情况，必须就当前操作在C和A之间做出选择。
+
+对于namingserver而言，我们更倾向于使用AP模型，即注重可用性和分区容错性，牺牲一定的一致性。NamingServer作为服务注册中心，主要职责是提供高效的服务发现与注册服务，而对短时间内的数据一致性要求可以适当放松。在分布式环境中，可能会出现多个节点短暂的注册数据不一致现象。例如，当多个NamingServer节点发生了网络分区时，某些节点获取的注册信息可能有延迟。
+对于NamingServer来说，这种短暂的不一致性我们认为是可以容忍的。由于服务注册与发现的强一致性要求并不高，即使在某个时刻有部分节点的注册数据滞后或不一致，也不会立刻影响到整个系统的正常服务。通过心跳机制、周期性同步等方式，最终一致性可以逐渐得到保证。
+
+
+
+### 3.2 Quorum NWR机制在NamingServer中的应用
+Quorum NWR（仲裁读写）是一种在分布式系统中用于确保数据一致性的机制。该机制通过设置副本的总数（N）、写操作需要成功的副本数（W）、读操作需要访问的副本数（R）来协调数据的一致性。在NamingServer的设计中，采用了多写+补偿机制来保证多个NamingServer节点之间的信息一致性，而客户端则与某一个NamingServer节点交互以获取注册信息。
+
+1. 写入操作（W-写仲裁）：
+   当有集群节点变化时，server端会将请求发送到NamingServer集群中的多个节点。
+   根据NWR机制，系统确保至少有W个副本成功写入注册信息。 通过多写机制，即使某些节点暂时不可用或存在网络延迟，仍能确保写操作的高可用性。一旦W个节点写入成功，客户端将收到成功响应。
+   补偿机制：对于没有立即成功写入的副本，系统会通过异步补偿方式，在稍后的时间段内同步这些节点，确保最终一致性。
+2. 读取操作（R-读仲裁）：
+   客户端通过与NamingServer集群中的任意一个节点进行交互，发送读取请求以获取服务注册信息。
+   系统会从至少R个副本中读取数据，采用最新版本的数据作为返回结果。即使某些节点的数据存在短暂不一致，客户端通过读取多个副本并比较其版本号或一致性标记，能够确保读取到最新的注册信息。
+   由于客户端只与一个NamingServer节点进行交互，读取操作的效率得到了提升，并且避免了多个节点之间的复杂协调，系统依然能保证最终一致性。
+3. NWR参数的设计与权衡：
+   在namingserver中我们设置W=N、R=1。W=N虽然意味着写入需要发送到所有节点，但并不要求所有节点必须立即成功写入。系统允许某些节点暂时失败，通过补偿机制在后续阶段同步这些节点，从而提高系统的容错性，因为即便某些节点在写入时发生故障或网络中断，数据更新仍然可以通过补偿机制最终传播到所有节点。这既保证了系统的高可用性，也确保了数据最终在所有节点上是一致的。由于写操作要求所有节点参与，因此每个节点都会接收到最新的数据更新。
+   客户端在执行读取操作时，可以从任意一个NamingServer节点读取数据，而不必担心数据不一致的问题。即使某些节点未能在写入时立即成功，客户端仍能从其他已成功写入的节点获取最新的注册信息。这样，R值可以设定较低（如R=1），从而提高读取操作的效率，同时系统通过补偿机制确保所有节点最终达到一致。
+![img](/img/blog/namingserver2.png)
+
+## 3.2 架构图
+
+
+![img](/img/blog/namingserver3.png)
+
+namingserver的运行链路如上图所示：
+
+1. 通过控制台在某个cluster下创建一个事务分组vgroup。
+2. 创建vgroup->cluster的请求发送给namingserver，namingserver再传递给对应的tc节点。
+3. tc节点将vgroup->cluster映射关系持久化保存。
+4. tc节点在心跳的时候将vgroup->cluster的映射关系更新到所有的namingserver。
+5. client通过自己配置的事务分组vgroup从namingserver获取对应cluster元数据。
+6. client在事务流程中，使用cluster下的unit进行负载均衡，再进行begin，registry，commit，rollback等。
+7. 事务决议后，对应的unit下的leader节点下发二阶段，无状态节点下，每个unit的唯一node就是leader。
+
+
+## 3.3 设计细节
+
+### 3.3.1 长轮询推送集群变化情况
+
+![img](https://camo.githubusercontent.com/82b14f95feb7ce17af5ca953dbb0340f1a86519c9067433944ac5fa8c26711c8/68747470733a2f2f63646e2e6e6c61726b2e636f6d2f79757175652f302f323032332f706e672f33333531373535322f313639353236343333343035342d38393065623863382d653437302d343761362d623030372d3732343365303938313732652e706e67)
+
+如上图所示，每隔30s client侧需要向namingserver发起一次服务发现的请求，用以拉取最新的tc列表。而在这30s的间隔中，client侧将采用HTTP长轮询的方式一直watch namingserver节点，如果namingserver 侧有如下的变化：
+
+> 1.事务分组映射关系的变化；
+>
+> 2.集群中实例的增加或者减少；
+>
+> 3.集群中实例的属性的变化；
+
+那么watch返回200状态码，告知client需要获取最新集群信息；否则namingserver将一直挂起watch方法，直到HTTP长轮询超时，然后返回304状态码， 告知client进行下一轮watch。
+