Mastering Ubuntu Server
上QQ阅读APP看书,第一时间看更新

Configuring administrator access with sudo

By now, we've already used sudo quite a few times in this book. At this point, you should already be aware of the fact that sudo allows you to execute commands as if you were logged in as root. However, we haven't had any formal discussion about it, yet nor have we discussed how to actually modify which of your user accounts are able to utilize sudo.

On all Linux systems, you should protect your root account with a strong password and limit it to be used by as few people as possible. Using sudo is an alternative to using root, so you can give your administrators access to perform root tasks with sudo without actually giving them your root password. In fact, sudo allows you to be a bit more granular. Using root is basically all or nothing—if someone knows the root password and the root account is enabled, that person is not limited and can do whatever they want. With sudo, that can also be true, but you can actually restrict some users to use only particular commands with sudo and therefore limit the scope of what they are able to do. For example, you could give an admin access to install software updates but not allow them to reboot the server.

By default, members of the sudo group are able to use sudo without any restrictions. Basically, members of this group can do anything root can do (which is everything). During installation, the user account you created was made a member of sudo. To give additional users access to sudo, all you would need to do is add them to the sudo group:

# usermod -aG sudo <username>

Note

Not all distributions utilize the sudo group by default, or even automatically install sudo. Other distributions require you to install sudo manually and may use another group (such as wheel) to govern access to sudo.

But again, that gives those users access to everything, and that may or may not be what you want. To actually configure sudo, we use the visudo command. This command assists you with editing /etc/sudoers, which is the configuration file that governs sudo access. Although you can edit /etc/sudoers yourself with a text editor, configuring sudo in that way is strongly discouraged. The visudo command checks to make sure your changes follow the correct syntax and helps prevent you from accidentally destroying the file. This is a very good thing, because if you did make some errors in the /etc/sudoers file, you may wind up in a situation where no one is able to gain administrative control over the server. And while there are ways to recover from such a mistake, it's certainly not a very pleasant situation to find yourself in! So, the takeaway here it to never edit the /etc/sudoers file directly; always use visudo to do so.

The following output is what I receive when I make an error while using visudo:

>>> /etc/sudoers: syntax error near line 32 <<<
What now?
Options are:
 (e)dit sudoers file again
 e(x)it without saving changes to sudoers file
 (Q)uit and save changes to sudoers file (DANGER!)

Thankfully, I used visudo and didn't edit that file manually! Otherwise, if I had edited the file manually, I may not have known that I made an error. You shouldn't necessarily rely on visudo catching your mistakes, but it is a nice safety net, just in case.

When you run visudo from your shell, you are brought into a text editor with the /etc/sudoers file opened up. You can then make changes to the file and save it like you would any other text file. By default, Ubuntu Server opens up the nano text editor when you use visudo. With nano, you can save changes using Ctrl + W, and you can exit the text editor with Ctrl + X. However, you don't have to use nano if you prefer something else. If you fancy the vim text editor, you can use the following command to use visudo with vim:

# EDITOR=vim visudo

Anyway, so visudo allows you to make changes to who is able to access sudo. But how do you actually make these changes? Go ahead and scroll through the /etc/sudoers file that visudo opens and you should see a line similar to the following:

%sudo   ALL=(ALL:ALL) ALL

This is the line of configuration that enables sudo access to anyone who is a member of the sudo group. You can change the group name to any that you'd like, for example, perhaps you'd like to create a group called admins instead. If you do change this, make sure that you actually create that group and add yourself and your staff to be members of it before you log off; otherwise, it would be rather embarrassing if you found yourself locked out of administrator access to the server.

Of course, you don't have to enable access by group. You can actually call out a username instead. As an example of this, we also have the following line in the file:

root    ALL=(ALL:ALL) ALL

Here, we're calling out a username (in this case, root), but the rest of the line is the same as the one I pointed out before. While you can certainly copy this line and paste it one or more times (substituting root for a different username), to grant access to others, using the group approach is really the best way. It's easier to add and remove users from a group (such as the sudo group), rather than use visudo each time.

So, at this point, you're probably wondering what the options on /etc/sudoers configuration lines actually mean. So far, both examples used ALL=(ALL:ALL) All. In order to fully understand sudo, understanding the other fields is extremely important, so let's go through them (using the root line again as an example).

The first ALL means that root is able to use sudo from any terminal. The second ALL means that root can use sudo to impersonate any other user. The third ALL means that root can impersonate any other group. Finally, the last ALL refers to what commands this user is able to do; in this case, any command he or she wishes.

To help drive this home, I'll give some additional examples. Here's a hypothetical example:

charlie ALL=(ALL:ALL) /usr/sbin/reboot,/usr/sbin/shutdown

Here, we're allowing user charlie to execute the reboot and shutdown commands. If user charlie tries to do something else (such as install a package), he will receive an error message:

Sorry, user charlie is not allowed to execute '/usr/bin/apt-get install tmux' as root on ubuntu-server.

However, if charlie wants to use the reboot or shutdown commands on the server, he will be able to do so because we explicitly called out those commands while setting up this user's sudo access. We can limit this further by changing the first ALL to a machine name, in this case, ubuntu-server to reference the host name of the server I'm using for my examples. I've also changed the command that charlie is allowed to run:

charlie ubuntu-server=(ALL:ALL) /usr/bin/apt-get

Now, the charlie is only able to use apt-get. He can use apt-get update, apt-get dist-upgrade, and any other subset of apt-get. But if he tries to reboot the server, remove protected files, add users, or anything else we haven't explicitly set, he will be prevented from doing so.

We have another problem, though. We're allowing charlie to impersonate other users. This may not be completely terrible given the context of installing packages (impersonating another user would be useless unless that user also has access to install packages), but it's bad form to allow this unless we really need to. In this case, we could just remove the (ALL:ALL) from the line altogether to prevent charlie from using the -u option of sudo to run commands as other users:

charlie ubuntu-server= /usr/bin/apt-get

On the other hand, if we actually do want charlie to be able to impersonate other users (but only specific users), we can call out the username and group that charlie is allowed to act on behalf of by setting those values:

charlie ubuntu-server=(dscully:admins) ALL

In that example, charlie is able to run commands on behalf of user dscully and group admins.

Of course, there is much more to sudo than what I've mentioned in this section. Entire books could be written about sudo (and have been), but 99% of what you will need for your daily management of this tool involves how to add access to users, while being specific about what each user is able to do. As a best practice, use groups when you can (for example, you could have an apt group, reboot group, and so on) and be as specific as you can regarding who is able to do what. This way, you're not only able to keep the root account private (or even better, disabled), but you also have more accountability on your servers.