Microservice automation

etcd Introduce

etcd Is a highly consistent distributed key value (key-value) Storage , It provides a reliable way to store data that needs to be accessed by distributed systems or machine clusters . It can gracefully handle the leader election during network partition , Even in Machine failures can also be tolerated in the leader node .

etcd Yes, it is Go language-written , It has excellent cross platform support , Small binaries and powerful communities .etcd Communication between machines is through Raft Consensus algorithm processing .

Service discovery

Service discovery is also one of the most common problems in distributed systems , That is, processes or services in the same distributed cluster , How to find each other and establish a connection

Service discovery is to find out whether there are processes listening in the cluster udp or tcp port , And through the name you can find and connect .

To solve the problem of service discovery , We need to have three pillars , Be short of one cannot :

A strong consistency 、 Highly available service storage directory . be based on Raft Algorithm etcd By nature, it is such a highly consistent and available service storage directory .

A mechanism for registering services and monitoring the health status of services . Users can go to etcd Registration service in , And set the registered service key TTL, Keep the heartbeat of the service regularly to achieve the effect of monitoring the health status .

A mechanism for finding and connecting services . By means of etcd The services registered under the specified topic can also be found under the corresponding topic . To make sure the connection , We can deploy one... On every service machine Proxy Mode etcd, That way To ensure access etcd The services of the cluster can be connected to each other .

​

Raft Election algorithm  

A cluster is usually composed of two or more servers , Each server is a node . Database cluster 、 Management cluster ... The database cluster provides read and write functions , The management cluster provides management 、 Fault recovery and other functions .

For a cluster , The coordination and management of multiple nodes is very important . The master node realizes collaboration and management , The existence of the master node , It can ensure the orderly operation of other nodes , And the write data in the database cluster on each node The consistency of .

Consistency here means , The data is the same in each cluster node , There is no difference . The function of distributed election is to select a master node , It coordinates and manages other nodes , To ensure the orderly operation of the cluster and the exchange of data between nodes Uniformity .

Raft The algorithm is a typical majority voting algorithm , Its election mechanism is similar to the democratic voting mechanism in our daily life , The core idea is “ The minority is subordinate to the majority ”

Raft In the algorithm, , The node that gets the most votes becomes the master node .

use Raft Algorithmic election , The roles of cluster nodes are 3 Kind of ：

Leader: The primary node , There is only one... At the same time Leader, Responsible for coordinating and managing other nodes ;

Candidate: The candidate , Every node can be Candidate, Nodes can only be selected as new under this role Leader;

Follower: Leader Follower , You can't have an election .

Raft Election process , It can be divided into the following steps ：

1. On initialization , All nodes are Follower state .

2. When you start the election , The state of all nodes is determined by Follower Turn into Candidate, And send election requests to other nodes .

3. Other nodes according to the order of election requests received , Reply whether you agree to become the master . What needs to be noted here is , In every election round , A node can only cast one vote .

4. If the node initiating the election request gets more than half of the votes , Then it becomes the master node , Its state is transformed into Leader, The state of other nodes is determined by Candidate Reduced to Follower.Leader Node and Follower Nodes send messages regularly Heartbeat bag , To detect whether the master node is alive .

5. When Leader The term of office of the node has arrived , That is to say, when other servers start the next round of main selection cycle ,Leader The state of a node is determined by Leader Downgrade to Follower, Enter a new round of election

 

etcd The term ：

Raft：etcd The algorithm used to ensure the strong consistency of the distributed system .

Node： One Raft State machine instance .

Member： One etcd example . It manages a Node, And can provide services for client requests .

Cluster： By multiple Member To form something that can work together etcd colony .

Peer： To the same etcd Another one in the cluster Member Address of .

Client： towards etcd Cluster send HTTP Requested clients .

WAL： Pre written logs ,etcd Log format for persistent storage .

snapshot：etcd prevent WAL Snapshot set due to too many files , Storage etcd Data status .

Proxy：etcd A model of , by etcd The cluster provides reverse proxy services .

Leader( The leader )：Raft The nodes in the algorithm that process all the data submitted through the election .

Follower( follower )： The node of the election defeat is Raft Slave nodes in , Provide strong consistency guarantee for algorithm .

Candidate： When Follower Can't receive... After a certain period of time Leader My heart rate changes to Candidate Start Leader campaign for .

Term： A node becomes Leader The time period until the beginning of the next election , Is called a Term.

Index： Data item number .Raft Pass through Term and Index To locate data .

etcd install (centos)

Upload etcd Installation package ( The Internet download is too slow )

tar -zxvf etcd-v3.4.3-linux-amd64.tar.gz

 

Switch to etcd root directory , take etcd and etcdctl Binary files are copied to /usr/local/bin Directory so that the system can directly call etcd/etcdctl These two programs

cp etcd etcdctl /usr/local/bin

 

see etcd edition

etcd --version

Enter the command etcd, You can start a single node etcd service ,ctrl+c You can stop the service

 

xxxx-xx-xx xx:xx:xx.xxxxxx I | embed: name = default

xxxx-xx-xx xx:xx:xx.xxxxxx I | embed: data dir = default.etcd

xxxx-xx-xx xx:xx:xx.xxxxxx I | embed: member dir = default.etcd/member

xxxx-xx-xx xx:xx:xx.xxxxxx I | embed: heartbeat = 100ms

xxxx-xx-xx xx:xx:xx.xxxxxx I | embed: election = 1000ms

xxxx-xx-xx xx:xx:xx.xxxxxx I | embed: snapshot count = 100000

xxxx-xx-xx xx:xx:xx.xxxxxx I | embed: advertise client URLs = http://localhost:2379

xxxx-xx-xx xx:xx:xx.xxxxxx I | etcdserver: starting member 8e9e05c52164694d in cluster cdf818194e3a8c32

1.name Represents the node name , The default is default.

2.data-dir  The directory where logs and snapshots are kept , The default is the current working directory default.etcd/ Under the table of contents .

3. stay http://localhost:2380 Communicating with other nodes in the cluster .

4. stay http://localhost:2379 Provide client interaction .

5.heartbeat by 100ms, The function of this parameter is leader How often does it send a heartbeat to followers, The default value is 100ms.

6.election by 1000ms, The function of this parameter is the timeout for re voting , If follow No heartbeat packets were received at this time interval , Will trigger a new vote , The default is 1000ms.

7.snapshot count by 10000, This parameter is used to specify how many transactions are committed , Trigger snapshot capture to save to disk .

8. The cluster and each node will generate a uuid.

9. When it starts, it will run raft, Election out leader

Create a etcd service

establish etcd Related contents ( That is, the storage location of data files and configuration files )

mkdir -p /var/lib/etcd/ && mkdir -p /etc/etcd/

 

establish etcd The configuration file

Stand-alone version conf

# The name of the node
ETCD_NAME="etcd0"
# Specify the data file storage location
ETCD_DATA_DIR="/var/lib/etcd/"

vim /etc/etcd/etcd.conf

 

establish systemd The configuration file

Stand-alone version service

[Unit]
Description=Etcd Server
After=network.target
After=network-online.target
Wants=network-online.target

[Service]
User=root
Type=notify
WorkingDirectory=/var/lib/etcd/
## Modify... According to the actual situation EnvironmentFile and ExecStart These two parameter values
## 1.EnvironmentFile That is, the location of the configuration file , Be careful “-” It can't be less
EnvironmentFile=-/etc/etcd/etcd.conf
## 2.ExecStart namely etcd Start program location
ExecStart=/usr/local/bin/etcd
Restart=on-failure
LimitNOFILE=65536

[Install]
WantedBy=multi-user.target

vim /etc/systemd/system/etcd.service

 

start-up / stop it / see etcd service

systemctl daemon-reload && systemctl enable etcd && systemctl start etcd && systemctl status etcd

 

etcd Basic use

etcdctl Is a command line client , It can be done to etcd Test the service or manually modify the database content . in addition ,etcdctl And support HTTP API( Later on ).

etcdctl The supported commands are generally divided into database operation and non database operation

Help order

etcdctl -h

put Put in the data

etcdctl put /testdir/testkey "Hello world"

 

--sort Sort results

--consistent Send the request to the master , Ensure consistency of acquired content .

del Delete data

Empty data

etcdctl del / --prefix

 

Delete all /test Prefix node

etcdctl del /test --prefix

watch, Monitoring the change of a key value , Once the key value is updated , Will output the latest value and exit

etcdctl watch key1

 

etcd Foundation and use

High availability cluster