(This is a reprint of the article I wrote on my blog)

Postscript: 2020/4/17

@ hirosys-biz has published the procedure to automatically build an environment with CloudFormation. It also solves the "bad" part of this article, so be sure to check it out as well. The environment for participating in the analysis project of COVID-19 (SARS-CoV-2) was automatically constructed by CloudFormation

Introduction

"Let's practice building an Amazon Web Services environment while contributing to the analysis of the new coronavirus." As a result, I came to the conclusion that "it worked normally, but I couldn't contribute much because of the lack of specifications in the free tier." Keep a record of your work.

background

Corona fashion is spicy. Isn't it possible to solve it sooner?

↓

An article on PC Watch that says "How to use" Folding @ home ", which contributes to corona analysis that can be started in 3 minutes even now" Find

↓

Immediately install it on your home laptop (with an older generation GPU). I've been running for about a week and have done about 25 tasks, but I still feel unsatisfactory.

↓

I can't raise the specs, so I want to increase the number of units in operation. But I only have one physical machine ...

↓

Yes, I had an AWS account. Let's use this. If you can do it in the free frame, can someone imitate it if you make it an article?

↓

Action.

What I did ①: Creating a VPC environment

I decided to build a VPC environment first and then set up an EC2 environment in it. Basically, work according to this blog article.

Only the security group has been changed as follows according to the purpose of use this time.

Port 80 is closed because we do not set up a web service.
Instead, open 7396 for the Folding @ Home web controller
As a security measure, open SSH ports other than No. 22, and leave No. 22 for initial login.
This environment is "only you can access", so all sources are limited to my IP

What I did (2): Build an EC2 environment on a VPC

Create an EC2 environment on the VPC environment created in ①. This work also referred to this blog article.

This time I wanted to create an environment that can be created with a free usage frame, so the configuration is as follows

OS: Amazon Linux 2 AMI(x86)
Instance type: t2.micro
Storage: General purpose SSD 8GB

What I did ③: Linux settings

Change SSH port number

$ sudo vi /etc/ssh/sshd_config

###Add the following around 22 lines(Port number specified in ①)
Port 2****

###After saving
$ sudo service sshd reload

Create an account with an RSA key different from when EC2 was created, and delete the default "ec2-user"

###New user creation(root privileges)
# useradd newuser
# passwd newuser

###SSH settings(root privileges)
# cd /home/newuser/
# mkdir .ssh
# chmod 700 .ssh
# chown newuser:newuser .ssh
###Send the public key here with SCP
# mv /home/ec2-user/id_rsa.pub .ssh/authorized_keys
# chmod 600 .ssh/authorized_keys
# chown newuser:newuser .ssh/authorized_keys

###Give permission to sudo
# usermod -aG wheel newuser

###In another window"newuser"Log in as. After that, operate on the screen on the new user side
$ sudo su -
###After confirming that you can become root, ec2-user deleted
# userdel ec2-user

What I did ④: Introducing Folding @ Home

This time, only the minimum required FAH Client is introduced.

Installation

$ wget https://download.foldingathome.org/releases/public/release/fahclient/centos-6.7-64bit/v7.6/fahclient-7.6.9-1.x86_64.rpm
$ sudo rpm -i --nodeps fahclient-7.6.9-1.x86_64.rpm

Setting
Edit /etc/fahclient/config.xml. Root privilege is required for updating.

<config>
  <!-- Client Control -->
  <fold-anon v='true'/>

  <!-- Folding Slot Configuration -->
  <!-- v7.6.9-COVID for cause after 1-If you specify 19, corona related tasks will be assigned preferentially.-->
  <cause v='COVID_19'/>
  <gpu v='false'/>

  <!--Access authority setting to the Web controller. xxx.xxx.xxx.Enter your IP address in place of xxx-->
  <allow>127.0.0.1 xxx.xxx.xxx.xxx</allow>
  <web-allow>127.0.0.1 xxx.xxx.xxx.xxx</web-allow>

  <!--How much CPU resources do you use? light,middle,3 choices of full-->
  <power v='full'/>

  <!-- User Information -->
  <!--User information. If you enter the user name, how much you contributed individually will be aggregated-->
  <user v='deflat'/>
  <!--The ID of the team you want to belong to. Optional. By the way, 222 is"Japan"team.-->
  <team v='222'/>
  <!--Optional. You can get more points by registering your e-mail address on the official website and setting the received key.-->
  <passkey v='***Paste the key here***'/>

  <!--Slot number and type. This time there is only one CPU so don't mess with it-->
  <!-- Folding Slots -->
  <slot id='0' type='CPU'/>

</config>

Reflect settings & restart service
Even if you stop during task processing, the next time you start, it will calculate from the continuation properly.

$ sudo /etc/init.d/FAHClient stop
$ sudo /etc/init.d/FAHClient start

result

If you access http: // <IP address of EC2 environment>: 7396 with a web browser, you can see that it works properly as shown below. ... but ** insanely slow **
The image is about 24 hours after the start of processing. Even with a personal laptop, it should take 3 to 4 hours.

What I found out later

The t2 instance of EC2 has a concept of "CPU credit balance", and when it is dead, the CPU resources that can be used are limited.
The t2-micro used this time has a CPU usage rate of only 10%.
It seems that if you stop the instance itself → start it, it will recover to the initial credit, so I will consider it in the future.
[Reference](http://tagvote.grinspace.jp/aws%E3%81%AEt2-micro%E3%82%A4%E3%83%B3%E3%82%B9%E3%82%BF%E3 % 83% B3% E3% 82% B9% E3% 81% AB% E3% 81% A4% E3% 81% 84% E3% 81% A6% EF% BC% 88% E3% 81% 8A% E3% 81 % 95% E3% 82% 89% E3% 81% 84% EF% BC% 89 /)

Summary and impressions

"Let's practice building an Amazon Web Services environment while contributing to the analysis of the new coronavirus."
Even for the first time, I was able to complete the entire process in 2 to 3 hours. Thank you to all our ancestors.
In the future, I would like to try the method of periodically executing "Stop instance → Credit recovery by starting" as mentioned above.
In addition, I would like to try the same with other IaaS services.

Change log

2020/4/20: Compatible with the latest version of FAH Client (Ver 7.6.9-1)

The story of trying to contribute to COVID-19 analysis with AWS free tier and failing