the first time I had a machine provisioned using a set of scripts I noticed that I had a hardcoded value for the amount of swap space that it could handle.

Damn! All my machines are using the same amount of swap! Can I fix that?

illustration of an increase in the size of the swap in a Linux system

It turns out that increasing the swap space is pretty straightforward.

tl;dr: just create an extra file and activate it - Linux will add it to the count of available swap space.

If you’ve ever needed to extend the amount of swap space that you have in your Linux machine, be it Ubuntu, CentOS or any other distribution, this blog post is for you.


To simulate the action, we can follow some steps:

  1. Prepare a virtual machine where we can mess with swap space.
  2. Set up a single swap area and activate it
  3. Expand the total amount of swap by creating an extra file, setting it as a swap area and enabling it.

Once the simulation is done, we can verify that we can end up with a system that has more swap space than before.

1. Preparing a VM

The easiest way of creating a VM that I know is making use of the excellent Vagrant.

It allows you to prepare a ruby file that tells it how to create a set of machines given a spec that you define.

What I like the most of it is that even though it’s ruby (and not something like yaml) it’s very declarative (I know nothing about ruby and can go with it pretty well).

Having vagrant properly installed we need to create a Vagrantfile:

# -*- mode: ruby -*-
# vi: set ft=ruby :

# run this little script that changes
# the password of the ubuntu user after
# the machine gets booted up
$script = <<SCRIPT
echo "ubuntu:admin" | chpasswd

Vagrant.configure(2) do |config|
        # set the image to be used to be ubuntu/zesty64
        config.vm.box = "ubuntu/zesty64"
        # set the hostname of the machine
        config.vm.hostname = "test-machine"
        # do not check for base image updates
        config.vm.box_check_update = false
        # do not sync the default vagrant directory
        config.vm.synced_folder ".", "/vagrant", disabled: true
        # provision the machine with a custom script
        config.vm.provision "shell", inline: $script
        # configure some parameters from the virtualbox provider
        config.vm.provider "virtualbox" do |v|
                v.memory = 512
                v.cpus = 1

Running vagrant up gives us a virtual machine.

vagrant up
Bringing machine 'default' up with 'virtualbox' provider...
==> default: Importing base box 'ubuntu/zesty64'...
==> default: Matching MAC address for NAT networking...
    default: 22 (guest) => 2200 (host) (adapter 1)
==> default: Setting hostname...
==> default: Running provisioner: shell...
    default: Running: inline script

Now we can SSH into the machine at any time:

vagrant ssh default

# or

ssh ubuntu@localhost -p 2200

# password is `admin` as set by the provisioning
# script that we defined in the Vagrantfile

Welcome to Ubuntu 17.04 (GNU/Linux 4.10.0-26-generic x86_64)

 * Documentation:  https://help.ubuntu.com


Now that we have a VM running we can give it some swap space. All the following commands are meant to be executed inside the VM.

2. Activating some wrong swap space

The first thing to do is allocate some space on a disk that Linux can refer to when swapping. Here we can choose two approaches: fallocate or dd.

fallocate is used to manipulate the allocated disk space for a file, either to deallocate or preallocate it. For filesystems which support the fallocate system call, preallocation is done quickly by allocating blocks and marking them as uninitialized, requiring no IO to the data blocks.

This is much faster than creating a file by filling it with zeroes as one would do with dd. You can check it with the following snippet:

time dd if=/dev/zero of=$swap_file bs=1M count=1024

1024+0 records in
1024+0 records out
1073741824 bytes (1.1 GB, 1.0 GiB) copied, 1.51977 s, 707 MB/s

real	0m1.534s
user	0m0.008s
sys	0m0.756s

Now comparing to the fallocate command:

rm -rf $swap_file
time fallocate --length 1g $swap_file;

real	0m0.022s
user	0m0.000s
sys	0m0.000s

ps.: this amount of time might not matter much if we’re allocating just a small amount like 1g and it’s not at instance-startup time.

However, having something like 64G could take some unnecessary time from the instance startup.

After space on disk has been reserved for swap we need now to make it so that Linux understands that file as a linux swap area designed for such purpose - mkswap is the command to do the job.

mkswap $swap_file

mkswap: /mnt/1g.swap: insecure permissions 0644, 0600 suggested.
Setting up swapspace version 1, size = 1024 MiB (1073737728 bytes)
no label, UUID=0ef91bd0-3691-4dde-848c-44699289d733

Once the operation succeeds, we need to activate the swap area. swapon is the command that does so. You just need to specify the file and Linux will keep a reference to it to swap in when needed.

free -h
              total        used        free      shared  buff/cache   available
Mem:           486M         62M        361M        1.8M         62M        406M
Swap:            0B          0B          0B
                || NO SWAP

swapon $swap_file

free -h
              total        used        free      shared  buff/cache   available
Mem:           486M         62M        360M        1.8M         62M        406M
Swap:          1.0G          0B        1.0G
                || SWAP ACTIVATED

Ok, we got some swap space! Now, can we make it bigger?

3. Extending it

Sure, making it bigger is just a matter of adding an extra file where we can submit swap memory to it and making sure that Linux keeps track of it.

Just execute the same commands as before but targetting a new file:

dd if=/dev/zero of=$swap_file bs=1M count=2048
mkswap $swap_file
swapon $swap_file

After all, we should now have 3G of swap space:

free -h
              total        used        free      shared  buff/cache   available
Mem:           486M         62M        7.1M        1.8M        416M        406M
Swap:          3.0G          0B        3.0G

That’s it!


Q: Can you allocate as much space as you want?

No, there’s an amount in the number of pages that Linux will allow to be addressed by swap area headers.

Q: Can you create as many swap areas as you want?

No, there’s a limit on that amount as well and each allocated swap area can be found in /proc/swaps.

Closing thoughts

This was one of those simple little things that you learn by doing and really has no secrets.

I already knew how to increase swap because I had previously hardcoded things before and wanted to change later but this one I also faced in the middle of a call when a machine was literally close to getting OOM triggered because all the RAM had been taken and swap space was almost done.

To have just a little bit more of time to investigate the leak I just had to increase the swap space and then let it crash after the investigation. These steps did the job.

By the way, if you want to know more about Linux in general, some tips regarding on-call stuff, operations, and software engineering you should definitely subscribe to the newsletter - I occasionally send some links and content that I’d really like to receive.

If you have any questions or just want to chat a little bit, let me know! I’m @cirowrc on Twitter.

Have a good one!