This is a container image for running Zulip (Github) in production. Image available from:
- Quay.io (
docker pull quay.io/galexrt/zulip:1.8.1-0
) - Docker Hub (
docker pull galexrt/zulip:1.8.1-0
)
Current Zulip version: 1.8.1
Current Docker image version: 1.8.1-0
This project defines a Docker image for a Zulip server, as well as
sample configuration to run that Zulip web/application server with
each of the major services that Zulip uses in
its own container: redis
, postgres
, rabbitmq
, memcached
.
We have configuration and documentation for Docker Compose and Kubernetes; contributions are welcome for documenting other container runtimes and flows.
If you aren't already a Docker expert, we recommend starting by reading our brief overview of how Docker and containers work in the next section.
Docker and other container systems are built around shareable
container images. An image is a read-only template with instructions
for creating a container. Often, an image is based on another image,
with a bit of additional customization. For example, Zulip's
zulip-postgresql
image extends the standard postgresql
image (by
installing a couple postgres
extensions). And the zulip
image is
built on top of a standard ubuntu
image, adding all the code for a Zulip
application/web server.
Every time you boot a container based on a given image, it's like booting off a CD-ROM: you get the exact same image (and anything written to the image's filesystem is lost). To handle persistent state that needs to persist after the Docker equivalent of a reboot or upgrades (like uploaded files or the Zulip database), container systems let you configure certain directories inside the container from the host system's filesystem.
For example, this project's docker-compose.yml
configuration file
specifies a set of volumes where
persistent Zulip data should be stored under
/opt/docker/zulip/
in the container host's file system:
/opt/docker/zulip/postgresql/data/
has the postgres container's persistent storage (i.e. the database)./opt/docker/zulip/zulip/
has the application server container's persistent storage, including the secrets file, uploaded files, etc.
This approach of mounting /opt/docker
into the container is the
right model if you're hosting your containers from a single host
server, which is how docker-compose
is intended to be used. If
you're using Kubernetes, Docker Swarm, or another cloud container
service, then these persistent storage volumes are typically
configured to be network block storage volumes (e.g. an Amazon EBS
volume) so that they can be mounted from any server within the
cluster.
What this means is that if you're using docker-zulip
in production
with docker-compose
, you'll want to configure your backup system to
do backups on the /opt/zulip/docker
directory, in order to ensure
you don't lose data.
To use docker-zulip
, you need the following:
- An installation of Docker and Docker Compose or a Kubernetes runtime engine.
- At least ~1GB of available RAM. For running the Zulip system (including databases, etc.) on a single VM not using Docker, we recommend at least 2GB of available RAM for running a production Zulip server. But if you're just testing and/or aren't expecting a lot of users/messages, you can get away with significantly less, because Docker makes it easy to sharply limit the RAM allocated to the services Zulip depends on, like redis, memcached, and postgresql (at the cost of potential performance issues).
To use this project, we recommend starting by cloning the repo (since
you'll want to edit the docker-compose.yml
file in this project):
git clone https://github.com/galexrt/docker-zulip.git
cd docker-zulip
# Edit `docker-compose.yml` to configure; see docs below
If you're in hurry to try Zulip, you can skip to start the Zulip server, but for production use, you'll need to do some configuration.
With docker-compose
, it is traditional to configure a service by
setting environment variables declared in the zulip -> environment
section of the docker-compose.yml
file; this image follows that
convention.
Mandatory settings. You must configure these settings (more discussion in the main Zulip installation docs):
SETTINGS_EXTERNAL_HOST
: The hostname your users will use to connect to your Zulip server. If you're testing on your laptop, you the default oflocalhost.localdomain
is great.SETTINGS_ZULIP_ADMINISTRATOR
: The email address to receive error and support emails generated by the Zulip server and its users.
Mandatory settings for serious use. Before you allow production traffic, you need to also set these:
POSTGRES_PASSWORD
andSECRETS_postgres_password
should both be a password for the Zulip container to authenticate to the Postgres container. Since you won't use this directly, you just want a long, randomly generated string. You can rotate these by just restarting the container.RABBITMQ_DEFAULT_PASS
andSECRETS_rabbitmq_password
are similar, just for the RabbitMQ container.SECRETS_secret_key
should be a long (e.g. 50 characters), random string. This value is important to keep secret and constant over time, since it is used to (among other things) sign login cookies.SETTINGS_EMAIL_*
: Where you configure Zulip's ability to send outgoing email.
Other settings. If an environment variable name doesn't start with
SETTINGS
or SECRETS
in docker-compose.yml
, it is specific to the
Docker environment. Standard Zulip server settings
are secrets are set using the following syntax:
SETTINGS_MY_SETTING
will becomeMY_SETTING
in/etc/zulip/settings.py
SECRETS_my_secret
will becomemy_secret
in/etc/zulip/zulip-secrets.conf
.
Reading the comments in the sample Zulip's settings.py file) is the best way to learn about the full set of Zulip's supported server-level settings.
Most settings in Zulip are just strings, but some are lists (etc.) which you need to encode in the YAML file. For example,
- For
AUTHENTICATION_BACKENDS
, you enterZULIP_AUTH_BACKENDS
as a comma-separated list of the backend names (E.g."EmailAuthBackend,GitHubAuthBackend"
).
SSL Certificates. By default, the image will generate a self-signed cert. We will soon also support certbot for this, just like we do in normal Zulip installations (contributions welcome!).
You can also provide an SSL certificate for your Zulip server by
putting it in /opt/docker/zulip/zulip/certs/
(by default, the
zulip
container startup script will generate a self-signed certificate and
install it in that directory).
The way the environment variables configuration process described in
the last section works is that the entrypoint.sh
script that runs
when the Docker image starts up will generate a
Zulip settings.py file file based on your settings
every time you boot the container. This is convenient, in that you
only need to edit the docker-compose.yml
file to configure your
Zulip server's settings.
An alternative approach is to set MANUAL_CONFIGURATION: "True"
and
LINK_SETTINGS_TO_DATA
in docker-compose.yml
. If you do that, you
can provide a settings.py
file and a zulip-secrets.conf
file in
/opt/docker/zulip/zulip/
, and the container will use those.
You can boot your Zulip installation with:
docker-compose up
This will boot the 5 containers declared in docker-compose.yml
. The
docker-compose
command will print a bunch of output, and then
eventually hang once everything is happily booted, usually ending with
a bunch of lines like this:
rabbitmq_1 | =INFO REPORT==== 27-May-2018::23:26:58 ===
rabbitmq_1 | accepting AMQP connection <0.534.0> (172.18.0.3:49504
-> 172.18.0.5:5672)
You can inspect what containers are running in another shell with
docker-compose ps
(remember to cd
into the docker-zulip
directory first).
If you hit Ctrl-C
, that will stop your Zulip server cluster. If
you'd prefer to have the containers run in the background, you can use
docker-compose up -d
.
You can now connect to your Zulip server. For example, if you set
this up on a laptop with the default port mappings and
SETTINGS_EXTERNAL_HOST
, typing http://localhost/
will take you to
your server. Note that in this default scenario, (1) you'll have to
proceed past a self-signed SSL error, and (2) you won't be able to
login until you create an organization, but visiting the URL is a good
way to confirm that your networking configuration is working
correctly.
You can now follow the normal instructions for how to create a Zulip organization and log in to your new Zulip server (though see the following section for how to run management commands).
From time to time, you'll need to attach a shell to the Zulip
container so that you can run manage.py
commands, check logs, etc.
The following are helpful examples:
# Get a (root) shell in the container so you can access logs
docker-compose exec zulip /bin/bash
# Create the initial Zulip organization
docker-compose exec zulip sudo -H -u zulip -g zulip /home/zulip/deployments/current/manage.py generate_realm_creation_link
Since that process for running management commands is a pain, we recommend using a wrapper script for running management commands.
A Kubernetes pod file is in the kubernetes/
folder; you can run it
with kubectl create -f ./kubernetes/
.
You should read the docker-compsoe
section above to understand how
this works, since it's a very similar setup. You'll want to to clone
this repository, and edit the zulip-rc.yml
to configure the image, etc.
The fastest way to get Kubernetes up and running for testing without signing up for a cloud service is to install Minikube on your system.
We are aware of two efforts at building Helm Charts for Zulip:
- A PR to the main Helm repo, which is further along.
- The zulip-helm project, which might be a helpful reference for work on this.
Contributions to finish either of those and get them integrated are very welcome! If you're interested in helping with this, post on this thread.
This image is not designed to make it easy to run multiple copies of
the zulip
application server container (and you need to know a lot
about Zulip to do this sort of thing successfully). If you're
interested in running a high-availablity Zulip installation, your best
bet is to get in touch with the Zulip support team at
support@zulipchat.com
.
When running your container in production, you may want to put your
Zulip container behind an HTTP proxy.
This wiki page documents how to do this correctly
with nginx
.
You can upgrade your Zulip installation to any newer version of Zulip
with the following instructions. At a high level, the strategy is to
download a new image, stop the zulip
container, and then boot it
back up with the new image. When the upgraded zulip
container boots
the first time, it will run the necessary database migrations with
manage.py migrate
.
If you ever find you need to downgrade your Zulip server, you'll need
to use manage.py migrate
to downgrade the database schema manually.
-
(Optional) Upgrading does not delete your data, but it's generally good practice to back up your Zulip data before upgrading to make switching back to the old version simple. You can find your docker data volumes by looking at the
volumes
lines indocker-compose.yml
e.g./opt/docker/zulip/postgresql/data/
. -
Pull the new image version, e.g. for
v1.8.1
run:docker pull quay.io/galexrt/zulip:v1.8.1
. -
Edit your
docker-compose.yml
to point to the new image version, e.g.:
zulip:
image: "quay.io/galexrt/zulip:1.8.1-0"
- You can execute the upgrade by running:
# Stops the old zulip container; this beings your downtime
docker-compose stop
# Boots the new zulip container; this ends your downtime
docker-compose up
# Deletes the old container images
docker-compose rm
That's it! Zulip is now running the updated version. You can confirm you're running the latest version by running:
docker exec -it YOUR_ZULIP_CONTAINER_ID_HERE su -- zulip -c "cat
/home/zulip/deployments/current/version.py"
(Replace YOUR_ZULIP_CONTAINER_ID_HERE
with your container id, which
you can find using docker ps
)
-
Edit
Dockerfile
in this directory to specify the Git repository and the commit to use asZULIP_GIT_URL
andZULIP_GIT_REF
(e.g.master
or1.8.1
or445932cc8613c77ced023125248c8b966b3b7528
). -
Edit your
docker-compose.yml
to comment out theimage
line, e.g.:
zulip:
# image: "quay.io/galexrt/zulip:1.8.1-0"
build:
context: .
- Run
docker-compose build zulip
to build a Zulip Docker image from the specified Git version.
Then stop and restart the container as described in the previous section.
Common issues include:
- Invalid configuration resulting in the
zulip
container not starting; checkdocker-compose ps
to see if it started, and then read the logs for the Zulip container to see why it failed. - A new Zulip setting not being passed through the Docker
entrypoint.sh script properly. If you
run into this sort of problem you can work around it by specifying a
ZULIP_CUSTOM_SETTINGS
with one setting per line below, but please report an issue so that we can fix this for everyone else.
You can get community support and tell the developers about your experiences using this project on #production-help on chat.zulip.org, the Zulip community server.
In late May 2018, we completed a complete rewrite of this project's documentation, so we'd love any and all feedback!
We love community contributions, and respond quickly to issues and PRs. Some particularly useful ways to contribute right now are:
- Contribute to this documentation by opening issues about what confused you or submitting pull requests!
- Reporting bugs or rough edges!
Huge thanks to everyone who has contributed. Special thanks to
Alexander Trost, who created
docker-zulip
and did a huge amount of the early work required to
make a high-quality Docker image for Zulip possible.