The idea is to have a way for an admins to SSH into a server without having to use passwords. This adds a level of security to your server setup. Without private keys you have to enter your user name and password. This can be less secure than generating SSH keys and adding your public key to SSH, plus with keys you don’t have to remember passwords.
First you need to generate the SSH keys. I prefer the ed25519 algorithm which is a newer one. You can get more info here.
The code to create an ed25519 ssh key in the current users .ssh directory will look like this.
ssh-keygen -f ~/.ssh/key-name -t ed25519
The -f flag tells ssh-keygen the name of the files you want to create. The above command would create key-name(private key) and key-name.pub(public) key, in the current users .ssh directory. The ~ is a Linux shortcut meaning /home/current_user/ so you don’t have to type all that.
The -t flag tells ssh-keygen what type of algorithm to use. If you don’t specify the -f flag and give the file a name, then both files are output in the current users .ssh directory as ed25519 and ed25519.pub
Here I won’t be doing much explaining, just listing links so people can read about this befuddling issue.
It often boggles my mind how differently you must write shell scripts vs the command line commands. It is often very inconsistent, I hate inconsistencies.
Basically Aliases within Linux shell are not recognized without some fancy hacky code, WOOHOO. So you will lose your mind if you are trying to set and use aliases in shell scripts.
You can set aliases all day long, but your scripts wont use them.
Yes you read that correctly. You can set aliases in your script, even right before you want to use it and Linux is like GTFO, I have no idea what that is. It even fails without a notice/error most times. What you can do is set a normal variable and use it in place of an alias though.
Say you had a script named do-this-thing.sh and it was located deep in a directory like /etc/directory/directory/directory/directory/do-echo “$yarnBin” > /etc/profile.d/server-alias.shthis-thing.sh
You could do the following in BASH
not_alias=/etc/directory/directory/directory/directory/do-this-thing.shalias not working inside bash shell script
bash not_alias
The above would execute the do-this-thing.sh file. You can also permanently set aliases in your Shell script. This is handy even if you can’t use the aliases in your script directly, you can use them in the terminal command line later.
To permanently set aliases alias when not working inside bash shell script place them in your .bashrc file for the user you are logged in This is usually located in /home/username/.bashrc or you can put it in the user profile .profile file or other places. How you do this varies by Linux shell.
Top of the list Debian/Ubuntu removes apt-key support and doesn’t tell anyone they did it, doesn’t give anyone a single hint as to what to do. No just remove/deprecate shit and don’t tell a single soul on earth. This kind of stupid makes me want to leave the industry entirely. I get so tired of messed up and undocumented shit that wastes hours and hours and hours of my time. Someone needs kicked for this.
I kept saying this over and over and looking everywhere for answers. I finally found this page through googling tons of terms and combos until I found it. I kept thinking my errors would be in the logs I set in my Nginx configs, or even in my PHP configs. But they were continually empty, I was going insane. I seriously blew up on twitter.
BLOWING UP
I thought Symfony was simply suppressing or not passing the errors to Nginx. So the logs are located at the place in the link above from documentation and not in the location you set in the Nginx configs or PHP configs.
By default, log entries are written to the var/log/dev.log file when you’re in the dev environment. In the prod environment, logs are written to var/log/prod.log
What made this confusing was, the docs don’t have a logical link flow when you are reading them trying to learn Symfony. You later find the SymfonyCasts which are better. But what makes it most confusing is in the docs about configuring Nginx, it even shows the following.
To me this was showing how to set the error log. This does nothing by the way, not unless nginx itself has an error I guess.
For months I’ve wondered WTF, where are my error logs. I kept putting my app in dev mode so I could debug it via the browser.
Don’t do that!!!!!
To be honest getting Symfony working with Nginx is a pain in the ballsocks. The reason is, Nginx doesn’t pass environmental variables through to php scripts like Apache does/can. If you want that kind of fancy feature you must hack nginx up and use some perl script or something similar. Otherwise with Nginx you must set the environment variables twice, once in nginx and once in shell.
Why would you do that?
Why two locations? Yeah this really angered me and blew my mind at first too. As mentioned above Nginx doesn’t have any easy way to pass the environment variables you set at the Linux server level. This is important with Symfony because you often need to run things like Doctrine on the command line.
So I was setting my Nginx Environmental variables, the app would see them just fine. I’d go to run doctrine or tests and BOOM missing environmental variables like WTF? Or I would set them in the Linux environment, view them with printenv load the app in the browser and Nginx didn’t pass the values to my script. It took a lot of googling to figure that out with lots of trial and failure. To make matters worse, you have to change the environmental variable names in order to run tests so that symfony loads them, otherwise it hides the values.
WTF is happening?
I then found out through experimenting that you had to set the variables for the command line in the Linux environment too. How to permanently set Linux environmental variablees covers how to do that. It’s easier to just Bash script or ansible the entire process with Hashicorp packer than to try to manually maintain it all, setting vars in two different places etc.
So for months I’ve been going insane trying to find my error logs. Today I found the error logs.
If you search you will find different answers to this. You can do this in multiple ways, here I will talk about 2 ways, single command and multiple commands.
First the idea is to switch from say root user to a named user you created or was created for you on your Linux server to run commands as not the root user. The reason you want to do this is so that everything isn’t owned by the root user. Or you are installing something like PHP Composer which barfs on you if you run it as root user.
You will see some saying to use su others saying to use sudo (some bs options etc.) You will also see really wrong answers on Stack. I have no idea why you would use sudo over su, you can google that. But I do know that su switches users. Here is an article goes into more detail of su vs sudo and when you use both.
Single command syntax
So the first way is to run a single command directly inline. If you are the root user you simply use su The syntax to do so is as follows:
su - username "commandToExecute [command options and arguments]"
It has been my experience that the Double ” Quotes are required or else the shell gets confused. You may be able to use single quotes if you don’t use any variables within the quotes.
Multiple commands syntax
To more easily issue multiple commands or long commands you need to use Linux heredoc syntax.
Heredoc uses <
su - $username <<SHT
cd $serverDir
php $composerFile install
SHT
Like I said you can use any Delimiter you want. It is tradition to use all caps for the word, it makes it easier to spot. The ending word (EOF here) has to have no spaces or words before it. You can list any number of commands within that syntax and all will be executed by the user.
NOTE: After the ending EOF the shell returns the user to whatever user you were/are logged in as before the lines of code. If you are logged in as root, you are returned to root. Also when you issue the su command you are moved out of the directory you are in. That is why I used cd to move back to the directory I needed to be in.
Way back before modern times, like 10 years ago. Javascript had a much funkier way of defining objects. It was called prototypical inheritance. This is still how Javascript works, the classes, modules etc. were all recent additions to the language to make it easier to work with. It is not a very fun way to program because it is like looking at a GIANT JSON more than a class with methods.
So what did/does prototype inheritance look like? Well this…
function JsCollection() {
this.jsObject = new Object();
}
JsCollection.prototype = {
constructor: JsCollection,
addNamedProperty: function (property, value) {
// only add the property if it doesn't exist, return true if it was created
//return false if it was not, to allow for testing before adding a new property
var returnBool = false;
if (!this.jsObject.hasOwnProperty(property)) {
returnBool = true;
this.jsObject[property] = value;
}
return returnBool;
},
getElementCount: function () {
var elementCount = 0;
//loop through the object and add to the count
for (var elem in this.jsObject) {
//only add to the value if it is part of collection
//https://developer.mozilla.org/en-US/docs/Web/JavaScript/Guide/Working_with_Objects
if(this.jsObject.hasOwnProperty(elem)){
elementCount++;
}
}
return elementCount;
},
getElementValue: function (property) {
var returnProp = null;
if (this.jsObject.hasOwnProperty(property)) {
returnProp = this.jsObject[property];
}
return returnProp;
},
removeNamedProperty: function (property) {
if (this.jsObject.hasOwnProperty(property)) {
delete this.jsObject[property];
}
},
changePropertyValue: function (property, value) {
if (this.jsObject.hasOwnProperty(property)) {
this.jsObject[property] = value;
}
},
getAllNamedProperties: function () {
return this.jsObject;
},
namedPropertyExists: function (property) {
var propExists = false;
if (this.jsObject.hasOwnProperty(property)) {
propExists = true;
}
return propExists;
}
};
//var objProps = obj1.getAllNamedProperties();
// how to loop through object properties
//for(var prop in objProps){
// console.log("Property is " + prop + ' Property value is ' + objProps[prop]);
//}
As you can see this is an object with functions in old fashioned Javascript syntax. This was too confusing of a syntax for most people, plus you had to learn the inner workings of Javascript and how prototypical inheritance works.
I won’t try to explain it here as it is pretty complicated to wrap your head around. Some videos may help better than an article.
Configuring php can be confusing. PHP uses multiple configuration files, but the main ones are both named php.ini. You see with php you can have separate configurations for the app and the command line aka cli.
Why would you do that?
This may sound stupid at first but it is due to the fact that your command line is a different environment from your app environment. In Linux for example each user of the cli has their own environmental variables. These variables get passed to the php cli. This can drive you insane on linux. You think you set the configurations correctly, app works, try command line and BOOM all messed up.
But I set the damn configs WTF?
The two main files are located at
/etc/php/7.4/cli/php.ini
/etc/php/7.4/fpm/php.ini
Yes they have the same exact name. Yes they contain the same thing. However, one is used for the command line (cli) and the other is for apps (fpm). So you must set the settings in both. If you wanted to use 1 file for both you might be able to remove/rename 1 file such as the cli, set the fpm file. Then use a symlink from the fpm file to the cli directory. It would require some testing but it will probably work.
P.S. The php configuration file is the largest you will ever in your life see. I’m guessing a few thousand lines, it feels like that, but there is a ton of commenting and documentation to help you complete with links.
One issue with creating SSH keys is there are so many ways to do it and no one tells you why they do what they do. A quick search will reveal almost everyone has their own way of doing it. If you are new to ssh keys I suggest you read this article really quick.
In the github docs they tell you how to create the ssh keys like this
ed25519 is basically the newest type of key, it is supposed to be the most secure
-C is for adding a comment to the key. This helps you identify it later in places like ~/.ssh/known_hosts, ~/.ssh/authorized_keys and when you use the command ssh-add -L which prints out your public key info
-t specifies the type of key. The above command tells ssh-keygen to create an ed25519 type of key more info
-f /home/akashicseer/tests/ssh/file_name This tells ssh-keygen where to put the file. If you don’t specify the name then it will use a default of something like id_ed25519 for private key and id_ed25519.pub for the public key. The code above will put the files named file_name (private key) and file_name.pub ( public key) in the folder /home/akashicseer/tests/ssh/ If you don’t specify the full path to the exact folder your keys will be put into your users home directory in the default .ssh location. On Ubuntu this is /home/username/.ssh/
NOTE: for ssh deploy keys, don’t specify a passphrase when you create them or you will have to manually enter it later when Packer or whatever you use runs your provisioning code. That means you won’t be able to automate if you enter a passphrase, because it will ask the terminal user to enter the phrase to do a git clone.
There are different types of ssh keys. If you don’t add the ed25519 part then a regular ssh key of type rsa is created, this is the default type of ssh key. Basically Github documentation is showing how to create a secure type of key to use with your code deployments. You will use this key to clone your repository to your server instances.
Creating the key is only half the battle. You must decide how you will create the key, especially if you are automating deployments. When automating deployments the process becomes very complicated.
First either you create the keys you need in the instance you are creating then use the github api to add them to the proper repo. Or you create the keys on a local development computer and use something like Hashicorp Packer to upload the files to the server instance during creation. The latter is the easiest way especially in automation of the infrastructure.
If you are creating your keys locally and using Packer to upload them, you will need to login to Github and go to the deploy keys section of the specific repo to add your public key. The public key is the one that ends in .pub usually. The easiest way to copy the key value is to use xclip which I mention in this article.
If fully automating the process and creating the key on the actual instance, you must remember to eventually remove older keys. Github lets you have like 50 keys per repo max. If your repo needs to be deployed to many instances, such as a microservice structure you can contact them to get added key abilities. You could also reuse the same key, but that would require keeping the private key in a repo as well which probably isn’t a very good idea at all, since ssh keys are the same as passwords basically.
Also remember this. If you are using deploy keys only to deploy by cloning the repo, then deleting the key after the clone is perfectly fine. You only need to use this key one time to clone (aka deploy ) your code, after this it is useless. You can and probably should create new SSH keys for deployment each time and remove them from Github after you deploy, then delete them from the server instance.
Unless you plan on keeping the same instance up and trying to pull from the repo etc. That is messy. Personally I’d prefer to use Packer to create new instances when I need to and redeploy. This has the added benefit that I can upgrade the instance with security etc., test it, add my app code, test it, then swap over after migrating the database and other files. This is like creating a clean slate every time.
You will also need to know how to add the keys to ssh-agent and use them, which I cover in this article.
SSL is a very important subject. All websites/apps should be using it. However the docs will leave you scratching your head saying WTF? So I am creating this long list of resources for anyone else who ever has to learn how to use it.
Articles
First here is a link to the docs – this will cause confusion as nothing tells you how to use the pieces together. So it is like looking into a box of legos and knowing it builds something but you don’t even have a picture as a hint. The best you can do is use the pieces to build something that doesn’t even resemble the original creation.
I came across this issue when automating infrastructure provisioning. I needed a way to pull the repository code for my app in the provisioning scripts. I didn’t want to use the ssh keys I have setup for the entire Github account due to security. I discovered that github has the ability for you to add per repository SSH keys, called Deploy keys. The docs totally left me in the dark. I had no idea how to do any of this so I had to spend days researching. I decided to write this article to save everyone else hours of time scouring the internet trying to figure out how the hell to do this.
Why use Deploy keys?
Why would you want to use Deploy keys? When automating infrastructure provisioning you don’t want to expose your personal SSH keys. These deploy keys are going to be used only for cloning a repo, you may be able to use them for other things I didn’t research that not my problem. LOL.
SSH keys when setup correctly, allow a higher level of security than user name and password. Many people are automating by scripting a user name and password, that is BAD. Also if you don’t set a passphrase for the SSH key it won’t prompt for it in the shell terminal. Normally you want your ssh key secured with a pass phrase, but for infrastructure automation we need no pass.
I won’t cover how to automate the infrastructure that would be a series of articles. What I want to cover is how in specific to use multiple SSH keys.
The syntax is wacky as it gets. First off when you are using GIT to pull/push/clone etc. from Github, git is using SSH underneath. So in order to use multiple SSH keys you actually configure SSH not GIT, but git reads the command you type and interacts with SSH on your behalf. Totally confusing. My first few hours were filled with a lot of WTF?
Wait. WTF?
First off the SSH config is stored in your users .ssh directory. On most Linux distributions that is in the user you are logged in as home directory. Basically /home/username/.ssh/ this directory will hold your SSH certs, known_hosts file, config file and others. The ssh config file is always named config and goes in the .ssh directory. If you are logged in as root it will be /root/.ssh/config. Many times when provisioning a server automatically the only user you have at first is root.
Example ssh config file should look like this
Host hostAlias
User git
Hostname github.com
IdentityFile=/root/.ssh/id_rsa
Yes it goes in a file exactly like that, no equals, no semicolons no quotes, just 1980’s YAML LOLOL. The most confusing setting above, which gets more confusing if you read the docs, is Host. Just think of it as Alias. I have no idea why it is even called Host instead of Alias. That threw me and so many others off. I kept putting the same value I had for Hostname. Hostname is the exact name of the host where your repo is, github.com for this example. Identity file is the private key file location.
Another thing to look at is git for the User. You might be able to use other names, but next I’ll show how the name part ties in.
To use the above setting to clone a repo for example you would type the following at the command line.
See the User git and the Host hostAlias. This looks so similar to the regular clone command. For example here is another one of my Github repositories a public one so you can play with this.
git@github.com:AkashicSeer/phphtml.git
This is the default to clone a repo. This has a default name for User of git and a default value for Host of github.com. I haven’t experimented enough yet but I am guessing you can change the name in the configs to anything you want such as Billy and use a command like : billy@github.com:AkashicSeer/phphtml.git
So back to the question how do you use multiple SSH/Deploy keys with Git and Github?
Like this
Host hostAlias
User git
Hostname github.com
IdentityFile=/root/.ssh/id_rsa
Host otherAlias
User git
Hostname github.com
IdentityFile=/root/.ssh/id_rsa_2
Host billy
User git
Hostname github.com
IdentityFile=/root/.ssh/id_rsa_3
Each IdentityFile must be a unique ssh or deploy key, they are the same thing, both are ssh keys.
Then to clone from each for example you would use the following for example.
The format is User@Host:repo-owner-name/actual-repo.git
The dot . I am putting at the end of the clone IS AWESOME. It tells Git to clone into the current directory and don’t use the name of the repo as the directory name. Basically just clone the damn repo into this damn folder. Without the dot it includes the repo name too. I often just want /opt/app-directory < code in that folder.
AND DON’T FORGET THE SECRET SAUCE
Don’t forget the secret sauce
Now that you have multiple SSH keys you must do some special magic to let SSH know about the keys. For each key you have to tell the ssh-agent it’s name. Basically when SSH does it’s thing your SSH client has to give a list of keys to the SSH agent on the server you are contacting. GIT uses SSH so you must tell SSH where the keys are for your github accounts.
To do this on linux you start the ssh-agent then you add the keys. It is a bit of a pain. First you must start the agent, then you add the key.
#start the agent on linux like so
eval `ssh-agent`
ssh-add /path/to/your/private/key
The value you give to ssh-add command should be the ones you used for your IdentityFile definitions. You will need to add each private key to the agent before it will work.
To test that your setup is working you can do the following and read the output. If there was a problem it will tell you, like it couldn’t find the key.
Running those commands will let you know if everything is configured properly.
BUT IT GETS FUNNER GUYS
The fun is just beginning
All of the 999 things above are still not enough if you want to automate this process. If you do all of the above and try automating the process, github will prompt you for a passphrase for an ssh key. It won’t be the deploy key either, NO why do that, that would be logical and make sense. What it wants is the passhprhase to the entire account, not the deploy key.
How to fix this?
And there is still, still more, you must chmod the .ssh directory to 600 such as
chmod -R 600 /home/user/.ssh
or where ever your ssh files are stored.
You may also need to do the following.
Create a dummy instance. On this instance issue the git clone command. When it asks for the passphrase enter the passphrase for the account that OWNS THE REPO, not the deploy key passphrase which should be empty.
This will add github to known_hosts file. Now use cat to output that info and copy it. You can’t use xclip like I mention in another article, no that’s not allowed for some no brain reason. Once you copy the code from known_hosts create another file on your system called known_hosts. You will need to upload this file along with the ssh deploy keys so that you are not prompted during automated clones.
If there is some sort of openssh setting or a way to do this automatically, I haven’t found it yet.
If you would like more information on how to create the ssh deploy keys themselves, read this article I wrote.
If you want more information about ssh checkout my list of resources here
It tells Git to clone into the current directory and don’t use the name of the repo as the directory name. Basically just clone the damn repo into this damn folder. Without the dot it includes the repo name too. I often just want /opt/app-directory < code in that folder.
