Making a Perforce Server With Docker
Purpose⌗
There’s not many Perforce docker images out there, and I haven’t found any that are maintained and well documented. There’s actually a pretty good reason for that - Perforce is messy. It leaves files all over the system, there’s no real option to constrain it to a specific area, and it makes some assumptions about how you want your system to be run that don’t really translate to a containerized environment very well.
Fortunately, because Docker is magic, we can remedy most of these things. By the end of this, you’ll have your own Perforce server running comfortably on your own infrastructure. We’ll be using the industry standard docker-compose
container orchestrator for this, but a similar configuration can be shipped to Docker Swarm and Kubernetes.
As this is targetted at people that probably aren’t running large scale clustering software, we’re going to stick with docker-compose
and focus on a single server.
What is Perforce?⌗
Perforce is a version control system geared towards game developers. It handles large files, binary files, and locks on those files better than most other version control systems. Other version control systems you may be familiar with are git
, subversion
(or svn
for short), and mercurial
(or hg
for short).
For most workloads, you’d want to choose git
. That statement is almost purely backed by my opinion, but git
is the most popular version control system for a reason. It’s fantastic for most things.
One thing git
doesn’t do well is handle large binary files. I’ll spare the gory details of this, but storing a frequently edited video file (such as a cutscene or animation for a video game) makes git
very unhappy. Hence, Perforce.
Prerequisites⌗
For this tutorial we’re going to assume that you have installed docker
and docker-compose
and have the docker
daemon running on your machine. This is normally the part where I would put simple instructions on how to do that, but Docker is reorganizing its packages and install processes and my instructions would be quickly outdated.
We’ll also assume that you’re familiar with very basic command line navigation. If you’re not, you probably shouldn’t be trying to administer systems and should consult with someone that does.
If you want to learn, I highly recommend going to linuxcommand.org. That’s where I learned basic terminal literacy, and you can go through everything you need to know in the span of a few hours.
Update 2021/12/122 - Marc Wilson from PCWDLD sent me an email with a Linux commandline cheatsheet that should be useful for people that want a quicker way to get some shell literacy. You can check it out here. Thanks Marc!!
That aside, let’s dig into it!
The problems with Perforce⌗
Perforce makes a lot of assumptions about how it’s going to be run. It assumes that it is going to be a daemon, managed manually by CLI interactions or an admin interface. That doesn’t translate well to Docker, where a “system” only exists to do one thing and its lifecycle is managed automatically.
It also places files all over the system. You can configure a data directory, but its database is initialized from where the start command is run instead of in a dedicated location and all non-volume files in a Docker container are ephemeral.
This is fairly typical behavior and works well for installing the server on a bare metal machine, but nowadays it is often desirable to isolate all of your services from each other.
Let’s dig into solving this problem.
Creating our service⌗
Let’s create a docker-compose
project. This translates to doing a few very simple operations.
$ mkdir perforce && cd perforce
$ touch docker-compose.yml
$ touch Dockerfile
Now we need to make some volumes. A volume
, to Docker, is a folder inside the container that is physically located outside of the container. That means we can have persistent data and configuration directories.
Let’s define our service here.
version: '3'
services:
perforce:
build:
context: .
dockerfile: Dockerfile
restart: unless-stopped
volumes:
#- ./p4dctl.conf.d:/etc/perforce/p4dctl.conf.d
- ./perforce-data:/perforce-data
- ./dbs:/dbs
environment:
- P4PORT=1666
- P4ROOT=/perforce-data
ports:
- 1666:1666
Here we’ve defined our service with three volumes - one for configuration, one for data, and one for what Perforce refers to as “databases.”
One of them, the configuration volume, has been commented out so that it does not mount yet. This is because we need to generate those files before running our server for the first time.
Let’s go and make those directories now.
$ mkdir p4dctl.conf.d
$ mkdir perforce-data
$ mkdir dbs
And that’s…actually it! Now we need to figure out what our image looks like.
Making our image⌗
Perforce has some good documentation for installing it on Ubuntu, and luckily for us, there’s an official Ubuntu docker image that we can base ours off of.
Let’s go fill out that Dockerfile
!
FROM ubuntu:focal
# Update our main system
RUN apt-get update
RUN apt-get dist-upgrade -y
# Get some dependencies for adding apt repositories
RUN apt-get install -y wget gnupg
# Add perforce repo
RUN wget -qO - https://package.perforce.com/perforce.pubkey | apt-key add -
RUN echo 'deb http://package.perforce.com/apt/ubuntu focal release' > /etc/apt/sources.list.d/perforce.list
RUN apt-get update
# Actually install it
RUN apt-get install -y helix-p4d
# Go into our directory, start Perforce, and view the log outputs
CMD chown -R perforce:perforce /perforce-data && cd /dbs && p4dctl start master && tail -F /perforce-data/logs/log
Aaaaaaaaand that’s that! With docker-compose
, volumes are not mounted until you reach the container entrypoint. The entrypoint, in this case, is defined with the CMD
directive. As such, we have to enter our dbs
directory there and no earlier.
Populating the configuration directory⌗
Perforce, in all its majesty, will create files in our etc
directory during installation that it requires to run its configuration script. Fortunately, we can grab those from our newly built image and stash them in our volume mount.
$ docker-compose run -T --rm perforce tar czf - -C /etc/perforce/p4dctl.conf.d . | tar xvzf - -C p4dctl.conf.d/
Now we just need to uncomment that last volume from our docker-compose
service definition to have those files mounted in the running container:
version: '3'
services:
perforce:
build:
context: .
dockerfile: Dockerfile
restart: unless-stopped
volumes:
- ./p4dctl.conf.d:/etc/perforce/p4dctl.conf.d
- ./perforce-data:/perforce-data
- ./dbs:/dbs
environment:
- P4PORT=1666
- P4ROOT=/perforce-data
ports:
- 1666:1666
Now we’re ready to actually configure our server! Perforce includes a nice little helper script for this.
$ docker-compose run --rm perforce /opt/perforce/sbin/configure-helix-p4d.sh
Configure the server name as master
, ensure that the server root is set to /perforce-data
, and that the port is set to 1666
. You’ll also set up superuser credentials for your server during this step.
Spinning it up⌗
Spinning up your new service is almost disappointingly easy. We just set it running, and leave it be.
$ docker-compose up --build -d
And that’s it! You’ve now got a Perforce server running on port 1666
on this machine. Go make the game of your dreams!
Troubleshooting⌗
A number of issues can potentially come up during this. Try spinning up your service without detaching it (docker-compose up --build
- note the lack of -d
in that command) to see the log output. If it’s running but detached, you can run docker-compose logs -f
to view the log output in real time.
“P4PORT is not set”⌗
If you get an error stating that P4PORT
is not set, that means that it is unable to access its data ddirectory. This is because Perforce has its own user ID and group ID that it runs under, and there is no simple way to change that without manually editing the configuration. The default UID and GID numbers are 101
, and you can change ownership of that directory by running the following command -
$ sudo chown -R 101:101 perforce-data
…although the latest revision of the Dockerfile
in this post does do that by itself. That, however, can cause the following problem -
“Cannot read perforce-data
” (or “can’t stat perforce-data
”)⌗
You’re running Docker as an unprivileged user and are unable to read the generated perforce data files. There are two options to resolve this, with the first being much easier -
- Run the service as root (
sudo docker-compose up --build -d
) - Add your user to a group with GID
101
You can also modify the perforce configuration file (master.conf
in your p4dctl.conf.d
directory) to set the Umask
and Owner
fields to something more compatible with your host system user layout, but that’s significantly more involved and will not be covered here. Perforce does validate that the Umask
and Owner
fields are correct, so if your service does not start after messing with that, you have done something wrong in your permissions configuration.
I do not personally use Perforce and devised this tutorial for a friend, but I have verified on multiple environments that it is working. If you find any errors or have any questions, always feel free to email me!