This is an automated email from the ASF dual-hosted git repository.
casion pushed a commit to branch dev
in repository https://gitbox.apache.org/repos/asf/linkis-website.git
The following commit(s) were added to refs/heads/dev by this push:
new 10dbad9f942 Optimize distributed deployment (#730)
10dbad9f942 is described below
commit 10dbad9f94226b409d0ef617a49f017712986694
Author: aiceflower <[email protected]>
AuthorDate: Sat Jul 22 18:18:05 2023 +0800
Optimize distributed deployment (#730)
---
docs/deployment/deploy-cluster.md | 48 ++++++++++++++++------
.../current/deployment/deploy-cluster.md | 46 ++++++++++++++++-----
2 files changed, 71 insertions(+), 23 deletions(-)
diff --git a/docs/deployment/deploy-cluster.md
b/docs/deployment/deploy-cluster.md
index c540d29a783..1e6ab05948b 100644
--- a/docs/deployment/deploy-cluster.md
+++ b/docs/deployment/deploy-cluster.md
@@ -89,24 +89,48 @@ Total memory: 300 people online at the same time * 1G
memory for a single engine
## 2. Process of distributed deployment
->The following is just a reference example, taking two servers as an example
for distributed deployment. At present, the one-click installation script does
not have good support for distributed deployment, and manual adjustment and
deployment are required.
+>All services of Linkis support distributed and multi-cluster deployment. It
is recommended to complete stand-alone deployment on one machine before
distributed deployment, and ensure the normal use of Linkis functions.
-If you have already successfully deployed linkis in a stand-alone mode on
server A, and now you want to add a server B for distributed deployment, you
can refer to the following steps
+At present, the one-click installation script does not have good support for
distributed deployment, and manual adjustment and deployment are required. For
the specific distributed deployment, you can refer to the following steps,
assuming that the user has completed the single-machine deployment on machine A.
-Mode: Eureka service multi-active deployment, some services are deployed on
server A, and some services are deployed on server B
-### 2.1 Environment preparation for distributed deployment
+### 2.1 Environment preparation for distributed deployment
Like server A, server B needs basic environment preparation, please refer to
[Linkis environment
preparation](deploy-quick#3-linkis%E7%8E%AF%E5%A2%83%E5%87%86%E5% A4%87)
-### 2.2 Eureka multi-active configuration adjustment
-The registration center Eureka service needs to be deployed on server A and
server B,
+**Network Check**
+
+Check whether the service machines that need distributed deployment are
connected to each other, and you can use the ping command to check
+```
+ping IP
+```
+
+**Permission check**
+
+Check whether there is a hadoop user on each machine and whether the hadoop
user has sudo authority.
+**Required Environmental Checks**
+
+Each linkis service depends on some basic environments before starting or when
tasks are executed. Please check the basic environment of each machine
according to the table below. For specific inspection methods, refer to [Linkis
environment preparation] (deploy-quick#3-linkis%E7%8E%AF%E5
%A2%83%E5%87%86%E5%A4%87)
+
+|Service Name|Dependency Environment|
+|-|-|
+|mg-eureka|Java|
+|mg-gateway|Java|
+|ps-publicservice|Java、Hadoop|
+|cg-linkismanager|Java|
+|cg-entrance|Java|
+|cg-engineconnmanager|Java, Hive, Spark, Python, Shell|
+
+
+Note: If you need to use other non-default engines, you also need to check
whether the environment of the corresponding engine on the machine where the
cg-engineconnmanager service is located is OK. The engine environment can refer
to each [engine in use](https://linkis.apache.org/zh-
CN/docs/latest/engine-usage/overview) to check the pre-work.
+
+### 2.2 Eureka multi-active configuration adjustment
-Modify the Eureka configuration file, add the configuration addresses of both
Eurekas, and let the Eureka services register with each other.
-On server A, make the following configuration changes
+Modify the Eureka configuration file on machine A, add the Eureka
configuration addresses of all machines, and let the Eureka services register
with each other.
+On server A, make the following configuration changes, taking two Eureka
clusters as an example.
```
-Revise
$LINKIS_HOME/conf/application-eureka.yml和$LINKIS_HOME/conf/application-linkis.yml
configuration
+Modify $LINKIS_HOME/conf/application-eureka.yml and
$LINKIS_HOME/conf/application-linkis.yml configuration
eureka:
client:
@@ -120,11 +144,11 @@
wds.linkis.eureka.defaultZone=http:/eurekaIp1:port1/eureka/,http:/eurekaIp2:port
```
### 2.3 Synchronization of installation materials
-On server A, package the successfully installed directory `$LINKIS_HOME` of
linkis, then copy and decompress it to the same directory on server B.
-At this point, if the `sbin/linkis-start-all.sh` command is started on server
A and server B to start all services, then all services have two instances. You
can visit the eureka service display page `http:/eurekaIp1:port1, or
http:/eurekaIp2:port2` to view
+Create the same directory `$LINKIS_HOME` on all other machines as on machine
A. On server A, package the successfully installed directory `$LINKIS_HOME` of
linkis, then copy and decompress it to the same directory on other machines.
+At this point, if the `sbin/linkis-start-all.sh` script is executed to start
all services on server A and other machines, then all services have n
instances, where n is the number of machines. You can visit the eureka service
display page `http:/eurekaIp1:port1, or http:/eurekaIp2:port2` to view.
### 2.4 Adjust startup script
-According to the actual situation, determine the services that need to be
deployed on server A and server B,
+According to the actual situation, determine the Linkis service that needs to
be deployed on each machine,
For example, the microservice `linkis-cg-engineconnmanager` will not be
deployed on server A,
Then modify the one-click start-stop script of server A,
`sbin/linkis-start-all.sh`, `sbin/linkis-stop-all.sh`, and comment out the
start-stop commands related to the `cg-engineconnmanager` service
```html
diff --git
a/i18n/zh-CN/docusaurus-plugin-content-docs/current/deployment/deploy-cluster.md
b/i18n/zh-CN/docusaurus-plugin-content-docs/current/deployment/deploy-cluster.md
index 2eb72066035..ba837895e9c 100644
---
a/i18n/zh-CN/docusaurus-plugin-content-docs/current/deployment/deploy-cluster.md
+++
b/i18n/zh-CN/docusaurus-plugin-content-docs/current/deployment/deploy-cluster.md
@@ -91,21 +91,45 @@ EngineConnManager（ECM）使用总资源
## 2 分布式部署的流程
->以下只是一个参考样例，以两台服务器为例进行分布式部署。目前一键安装脚本对分布式部署还没有很好的支持，需要手动进行调整部署。
+>Linkis的所有服务都支持分布式、多集群部署，建议在分布式部署前，先在一台机器上完成单机部署，并保证 Linkis 各功能正常使用。
-假如已经在服务器A上，成功以单机方式部署了linkis，现在想添加一台服务器B，进行分布式部署，可以参考以下步骤
+目前一键安装脚本对分布式部署还没有很好的支持，需要手动进行调整部署。具体分布式部署，可以参考以下步骤，假设用户已经在 A 机器上完成了单机部署。
-模式：Eureka服务多活部署 ，部分服务部署在服务器A，部分服务部署在服务器B上
### 2.1 分布式部署的环境准备
和服务器A一样，服务器B需要进行基础的环境准备，请参考[Linkis环境准备](deploy-quick#3-linkis%E7%8E%AF%E5%A2%83%E5%87%86%E5%A4%87)
-### 2.2 Eureka多活配置调整
-注册中心Eureka服务，需要部署在服务器A和服务器B上，
+**网络检查**
+
+检查需要分布式部署的各服务机器是否网络互通，可用 ping 命令进行检查
+```
+ping IP
+```
+
+**权限检查**
+
+检查各机器上是否有 hadoop 用户以及hadoop用户是否有 sudo 权限。
+**必要的环境检查**
+
+Linkis各服务启动前或任务执行时会依赖一些基础环境，请根据下表检查各机器基础环境，具体检查方法参考[Linkis环境准备](deploy-quick#3-linkis%E7%8E%AF%E5%A2%83%E5%87%86%E5%A4%87)
+
+|服务名称|依赖环境|
+|-|-|
+|mg-eureka|Java|
+|mg-gateway|Java|
+|ps-publicservice|Java、Hadoop|
+|cg-linkismanager|Java|
+|cg-entrance|Java|
+|cg-engineconnmanager|Java、Hive、Spark、Python、Shell|
+
+
+注意：如需使用其它非默认引擎，还需要检查cg-engineconnmanager服务所在机器上对应引擎的环境是否OK，引擎环境可参考各[引擎使用中](https://linkis.apache.org/zh-CN/docs/latest/engine-usage/overview)中的前置工作进行检查。
+
+### 2.2 Eureka多活配置调整
-修改Eureka配置文件,把两台Eureka的配置地址都加上，让Eureka服务之间相互注册。
-在服务器A上，进行如下配置修改
+修改 A 机器上 Eureka 配置文件,把所有机器的 Eureka 的配置地址都加上，让 Eureka 服务之间相互注册。
+在服务器A上，进行如下配置修改，以两台 Eureka 集群为例。
```
修改
$LINKIS_HOME/conf/application-eureka.yml和$LINKIS_HOME/conf/application-linkis.yml配置
@@ -122,11 +146,11 @@
wds.linkis.eureka.defaultZone=http:/eurekaIp1:port1/eureka/,http:/eurekaIp2:port
```
### 2.3 安装物料的同步
-在服务器A上，将linkis的成功安装的目录`$LINKIS_HOME` 打包，然后拷贝并解压到服务器B的相同目录下。
-此时，如果在服务器A上以及服务器B上，启动`sbin/linkis-start-all.sh`命令启动所有服务，那么所有服务都有两个实例。
可以访问eureka服务展示页面 `http:/eurekaIp1:port1，或http:/eurekaIp2:port2` 查看
+在其它所有机器上创建同机器 A 上相同的目录 `$LINKIS_HOME` 。在服务器A上，将linkis的成功安装的目录`$LINKIS_HOME`
打包，然后拷贝并解压到其它机器相同目录下。
+此时，如果在服务器A上以及其它机器上，执行`sbin/linkis-start-all.sh`脚本启动所有服务，那么所有服务都有n个实例，n为机器数量。
可以访问eureka服务展示页面 `http:/eurekaIp1:port1，或http:/eurekaIp2:port2` 查看
### 2.4 调整启动脚本
-根据实际情况，确定服务器A和服务器B上需要部署的服务，
+根据实际情况，确定各机器上需要部署的Linkis服务，
比如 微服务`linkis-cg-engineconnmanager` 不会部署在服务器A上，
则修改服务器A的一键启停脚本，`sbin/linkis-start-all.sh`，`sbin/linkis-stop-all.sh`，将`cg-engineconnmanager`服务相关的启停命令，注释掉
```html
@@ -149,4 +173,4 @@ sbin/linkis-stop-all.sh
- 分部署部署时，linkis的安装目录建议保持一致，方便统一管控，相关的配置文件最好也保持一致
- 如果某些服务器，端口已被其他应用占用，无法使用时，需要调整该服务端口
- mg-gateway的多活部署，目前因为登陆session不支持分布式，用一个用户的请求，需要请求到同一个gateway实例上，可以通过nginx的ip
hash负载均衡方式来支持
-- 一键启停脚本，根据实际情况，进行调整，对于不再本本服务器上部署的微服务，需要在一键启动脚本中，将对应的启停命令注释掉。
\ No newline at end of file
+- 一键启停脚本，根据实际情况，进行调整，对于不再本服务器上部署的微服务，需要在一键启动脚本中，将对应的启停命令注释掉。
\ No newline at end of file
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
