Shulou(Shulou.com)06/02 Report--

Cluster and storage

Brief introduction (what is a cluster)

A group of computer groups interconnected through a high-speed network and managed in a single system mode

Bring many servers together to provide a service that looks like there is only one server on the client

Relatively high benefits in terms of performance, reliability and flexibility can be achieved at a lower cost.

Task scheduling is the core technology in cluster system.

Cluster purpose:

Improve performance: compute-intensive applications such as weather forecasting, nuclear test simulation

Reduce costs: cheaper than million-dollar supercomputers

Improve scalability: as long as you add cluster nodes

Enhanced reliability: multiple nodes perform the same function to avoid single point of failure

Cluster classification:

High availability Computing Cluster HPC

Solve complex scientific problems through parallel applications developed in clusters

Application: scientific calculation in weather, aviation, navigation and other scientific fields

Load balancing (LB) cluster

The client load is distributed as evenly as possible in the computer cluster.

High availability (HA) cluster

Avoid a single point of failure. When a system fails, it can migrate quickly.

Implementation of LB: LVS software Haproxy software Nginx

Implementation of HA: keepalived software

Introduction to the LVS project:

Linux Virtual Server (LVS) was created by Zhang Wensong during his PhD at National University of Defense Science and Technology.

LVS can implement highly available and scalable network services such as web, Mail, Cache and Media.

The ultimate goal is to realize a server application cluster with high availability, high performance and low cost by using Linux operating system, LVS cluster and LVS cluster software.

LVS cluster composition

Front end: load balancing layer, consisting of one or more sets

Middle: server group layer, consisting of a set of servers that actually run application services

Ground side: a data sharing storage layer that provides a storage area for shared storage space

LVS terminology

Director Server: degree server

Distribute the load to Real server's server

Real server: real server

A server that really provides application services

VIP: virtual IP address

The delicate IP address published for users to visit

RIP: real IP address

Use IP addresses on cluster nodes

DIP: the IP address of the node server to which the scheduler connects

LVS working mode

VS/NAT

Virtual server realized by network address translation

VS/DR

Directly use routing technology to realize virtual server

Node server needs to be configured with VIP, note MAC address broadcast

VS/TUN

Usually realize the virtual server through the tunnel

When different computer rooms in different locations need to provide services, you need to set LVS to VS/TUN mode.

LVS load balancing scheduling algorithm. Currently, 10 scheduling algorithms have been implemented.

Polling (Round Robin)

Weighted polling (Weighted Round Robin)

Minimum connection (Least Connections)

Weighted least connection (Weighted Least Connections)

LVS experiment (polling):

1. Configure LVS/NAT cluster

Host role:

Real server: hosts 52 and 53

Dispenser: host 54

Client: host 50

Storage services: host 51

Preparation for the experiment:

Dispenser: host 54

Eth0: private network IP address 192.168.4.54

Eth0: public network Ip address: 192.168.2.54

Enable the routing and forwarding function of the kernel

Real server

Host 52 eth0 192.168.4.52

Host Gateway: 192.168.4.54

Host: 53 eth0 192.168.4.53

Gateway: 192.168.4.54

Turn on host 54 kernel routing forwarding function

Sysctl-a | grep ip_forward / / check whether route forwarding is enabled. It is enabled by default.

Vim / etc/sysctl.conf

Net.ipv4.ip_forward = 1 / / newly added

Real server:

The route command is used to display and set the network routing table in the Linux kernel, and the routes set by the route command are mainly static routes.

Host 52ax 53: add Gateway

Route add default gw 192.168.4.54

Host 50: add gateway

Route add default gw 192.168.2.54

Note: route del default gw 192.168.4.54 / / misconfigured to delete with this command

Route-n / / View the routing table

Configure dispenser host 54

Ipvsadm usage:

A

Add a virtual server

-t

Set the cluster address (VIP,Virtual IP)

S

Specify load scheduling algorithm

-a

Add a real server

-d

Delete a real server

-r

Specify real server (real server)

-m

Use NAT mode;-g and-I correspond to DR and TUN mode, respectively

-w

Set the weight for the node server. Default is 1.

-C

Clear all ipvsadm configurations

-S

Permanently save the configuration

-E

Change scheduling algorithm

-e

-Z

Clear (ipvsadm-Ln-- stats) counter

VIP address: the address at which services are provided.

Vip: 192.168.2.54

Rip: 192.168.4.52

192.168.4.53

1. Install the ipvsadm package

Yum-y install ipvsadm.x86_64

two。 Rr scheduling algorithm and polling algorithm are used to create virtual service; wrr

] # ipvsadm-A-t 192.168.2.54 80-s rr

3. Add server node (real server) with-m in NAT mode

] # ipvsadm-a-t 192.168.2.54 VRV 80-r 192.168.4.52 Rd 80-m

] # ipvsadm-a-t 192.168.2.54 virtual 80-r 192.168.4.53 virtual 80-m

4. Save ipvs configuration

Ipvsadm- S > / etc/sysconfig/ipvsadm-config

5. View IPVS

] # ipvsadm-Ln

-> RemoteAddress:Port Forward Weight ActiveConn InActConn

TCP 192.168.4.54:80 rr

-> 192.168.4.52 Masq 80 1 0

-> 192.168.4.53 Masq 80 1 0

Forward: distribution mode

ActiveConn: the number of active connections

Weight: weight. In load balancing, the higher the value is, the more times the distribution will be.

InActConn:

Verify:

Ipvsadm-Ln-- stats

Prot LocalAddress:Port Conns InPkts OutPkts InBytes OutBytes

TCP 192.168.2.54:80 5 30 20 1980 2495

-> 192.168.4.52 purl 80 2 12 8 792 998

-> 192.168.4.53 purl 80 3 18 12 1188 1497

Conns: number of requested connections

InPkts: incoming packets

OutPkts: outgoing packet

InBytes: number of bytes of data coming in

OutBytes: the number of bytes of data coming in and out

Watch-n 1 ipvsadm-Ln-- stats / / watch-n 1 command refresh subsequent commands in 1 second

Manage the cluster

Delete the configuration of nat mode

] # ipvsadm-d-t 192.168.2.54 80-r 192.168.4.53

Add realserver (NAT)

] # ipvsadm-a-t 192.168.2.54 80-r 192.168.4.53-m

Modify scheduling algorithm

] # ipvsadm-E-t 192.168.2.54 80-s wrr

Modify the weight of real server

] # ipvsadm-e-t 192.168.2.54 80-r 192.168.4.53-w 2-m

Empty counter (ipvsadm-Ln-count of stats)

] # ipvsadm-Z

Clear all configurations

] # ipvsadm-C

Save all configurations

] # ipvsadm- S > / etc/sysconfig/ipvsadm-config

Restart the ipvsadm service

] # ipvsadm-S > / etc/sysconfig/ipvsadm

] # systemctl restart ipvsadm.service

Configure LVS/DR mode cluster

Principle: the request traffic is sent to the realserver by the dispatcher, and the response traffic is sent directly to the client by the server

Prepare:

Requirements: client access VIP address: 192.168.4.253 access website cluster

Configure dispenser 54

1. Bind the vip address 192.168.4.253 to the eth0 interface of this machine. Here, the command line is set to expire after a valid restart, and it will be sandwiched into the / etc/rc.load file to run automatically every time you boot.

] # ifconfig eth0:1 192.168.4.254/32

] # echo 'ifconfig eth0:1 192.168.4.254 echo 32' > / etc/rc.local

two。 Create a virtual service and use 192.168.4.253 as the VIP algorithm to use rr polling algorithm

] # ipvsadm-A-t 192.168.4.253 80-s rr

3. Add real server node,-g in DR mode

] # ipvsadm-a-t 192.168.4.253 virtual 80-r 192.168.4.52 virtual 80-g

] # ipvsadm-a-t 192.168.4.253 virtual 80-r 192.168.4.53 virtual 80-g

4. Save ipvs configuration

] # ipvsadm- S > / etc/sysconfig/ipvsadm-config

5. View IPVS

] # ipvsadm-Ln

-> RemoteAddress:Port Forward Weight ActiveConn InActConn

TCP 192.168.4.253:80 rr

-> 192.168.4.52 Route 80 1 0

-> 192.168.4.53 Route 80 1 0

Configure readlserver 52 and 53

1. Modify the kernel parameters of the network interface, hosts 52 and 53, to ensure that each boot will take effect later, append the following names to / etc/rc.load

Echo 1 > / proc/sys/net/ipv4/conf/lo/arp_ignore

Echo 2 > / proc/sys/net/ipv4/conf/lo/arp_announce

Echo 1 > / proc/sys/net/ipv4/conf/all/arp_ignore

Echo 2 > / proc/sys/net/ipv4/conf/all/arp_announce

two。 Bind the vip address 192.168.4.253 to 32 on the local lo interface, and add hosts 52 and 53.

Ifconfig lo:1 192.168.4.253/32

Echo 'ifconfig lo:1 192.168.4.253ax 32' > / etc/rc.local

3. Run the web service

Host 52:yum-y install httpd

Echo xiao > / var/www/html/index.html

Host 53:yum-y install httpd

Echo 123 > / var/www/html/index.html

4. Verify:

On the client

Curl http://192.168.4.253

Extended LVS:

When LVS service does LB cluster, it cannot do health line check on realserver service.

The solution is to periodically execute monitoring scripts on the dispenser

Vim / root/checklvs.sh

#! / bin/bash

How to know if realserver 80 is open can be checked with nmap

Nmap-sS-n-p 80 192.168.4.53 in this command to filter open if it is open, either open or close, give its value to a variable

Nmap-sS-n-p 80 192.168.4.53 | grep open

Execute the command to delete realserver locally when realserver 80 is not open

Ipvsadm-d-t 192.168.4.253 80-r 192.168.4.x

Add the command to delete realserver to the virtual service when realserver 80 is open

Ipvsadm-a-t 192.168.4.253 ipvsadm 80-r 192.168.4.x:80

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.