zulip/docs/production/deployment.md

29 KiB

Deployment options

The default Zulip installation instructions will install a complete Zulip server, with all of the services it needs, on a single machine.

For production deployment, however, it's common to want to do something more complicated. This page documents the options for doing so.

Installing Zulip from Git

To install a development version of Zulip from Git, just clone the Git repository from GitHub:

# First, install Git if you don't have it installed already
sudo apt install git
git clone https://github.com/zulip/zulip.git zulip-server-git

and then continue the normal installation instructions. You can also upgrade Zulip from Git.

The most common use case for this is upgrading to main to get a feature that hasn't made it into an official release yet (often support for a new base OS release). See upgrading to main for notes on how main works and the support story for it, and upgrading to future releases for notes on upgrading Zulip afterwards.

In particular, we are always very glad to investigate problems with installing Zulip from main; they are rare and help us ensure that our next major release has a reliable install experience.

Zulip in Docker

Zulip has an officially supported, experimental docker image. Please note that Zulip's normal installer has been extremely reliable for years, whereas the Docker image is new and has rough edges, so we recommend the normal installer unless you have a specific reason to prefer Docker.

Advanced installer options

The Zulip installer supports the following advanced installer options as well as those mentioned in the install documentation:

  • --postgresql-version: Sets the version of PostgreSQL that will be installed. We currently support PostgreSQL 10, 11, 12, 13, and 14.

  • --postgresql-database-name=exampledbname: With this option, you can customize the default database name. If you do not set this. The default database name will be zulip. This setting can only be set on the first install.

  • --postgresql-database-user=exampledbuser: With this option, you can customize the default database user. If you do not set this. The default database user will be zulip. This setting can only be set on the first install.

  • --postgresql-missing-dictionaries: Set postgresql.missing_dictionaries ([docs][doc-settings]) in the Zulip settings, which omits some configuration needed for full-text indexing. This should be used with cloud managed databases like RDS. This option conflicts with --no-overwrite-settings.

  • --no-init-db: This option instructs the installer to not do any database initialization. This should be used when you already have a Zulip database.

  • --no-overwrite-settings: This option preserves existing /etc/zulip configuration files.

Installing on an existing server

Zulip's installation process assumes it is the only application running on the server; though installing alongside other applications is not recommended, we do have some notes on the process.

Running Zulip's service dependencies on different machines

Zulip has full support for each top-level service living on its own machine.

You can configure remote servers for PostgreSQL, RabbitMQ, Redis, in /etc/zulip/settings.py; just search for the service name in that file and you'll find inline documentation in comments for how to configure it.

Since some of these services require some configuration on the node itself (e.g. installing our PostgreSQL extensions), we have designed the Puppet configuration that Zulip uses for installing and upgrading configuration to be completely modular.

For example, to install a Zulip Redis server on a machine, you can run the following after unpacking a Zulip production release tarball:

env PUPPET_CLASSES=zulip::profile::redis ./scripts/setup/install

All puppet modules under zulip::profile are allowed to be configured stand-alone on a host. You can see most likely manifests you might want to choose in the list of includes in the main manifest for the default all-in-one Zulip server, though it's also possible to subclass some of the lower-level manifests defined in that directory if you want to customize. A good example of doing this is in the zulip_ops Puppet configuration that we use as part of managing chat.zulip.org and zulip.com.

Using Zulip with Amazon RDS as the database

You can use DBaaS services like Amazon RDS for the Zulip database. The experience is slightly degraded, in that most DBaaS provides don't include useful dictionary files in their installations and don't provide a way to provide them yourself, resulting in a degraded full-text search experience around issues dictionary files are relevant (e.g. stemming).

You also need to pass some extra options to the Zulip installer in order to avoid it throwing an error when Zulip attempts to configure the database's dictionary files for full-text search; the details are below.

Step 1: Set up Zulip

Follow the standard instructions, with one change. When running the installer, pass the --no-init-db flag, e.g.:

sudo -s  # If not already root
./zulip-server-*/scripts/setup/install --certbot \
    --email=YOUR_EMAIL --hostname=YOUR_HOSTNAME \
    --no-init-db --postgresql-missing-dictionaries

The script also installs and starts PostgreSQL on the server by default. We don't need it, so run the following command to stop and disable the local PostgreSQL server.

sudo service postgresql stop
sudo update-rc.d postgresql disable

This complication will be removed in a future version.

Step 2: Create the PostgreSQL database

Access an administrative psql shell on your PostgreSQL database, and run the commands in scripts/setup/create-db.sql to:

  • Create a database called zulip.
  • Create a user called zulip.
  • Now log in with the zulip user to create a schema called zulip in the zulip database. You might have to grant create privileges first for the zulip user to do this.

Depending on how authentication works for your PostgreSQL installation, you may also need to set a password for the Zulip user, generate a client certificate, or similar; consult the documentation for your database provider for the available options.

Step 3: Configure Zulip to use the PostgreSQL database

In /etc/zulip/settings.py on your Zulip server, configure the following settings with details for how to connect to your PostgreSQL server. Your database provider should provide these details.

  • REMOTE_POSTGRES_HOST: Name or IP address of the PostgreSQL server.
  • REMOTE_POSTGRES_PORT: Port on the PostgreSQL server.
  • REMOTE_POSTGRES_SSLMODE: SSL Mode used to connect to the server.

If you're using password authentication, you should specify the password of the zulip user in /etc/zulip/zulip-secrets.conf as follows:

postgres_password = abcd1234

Now complete the installation by running the following commands.

# Ask Zulip installer to initialize the PostgreSQL database.
su zulip -c '/home/zulip/deployments/current/scripts/setup/initialize-database'

# And then generate a realm creation link:
su zulip -c '/home/zulip/deployments/current/manage.py generate_realm_creation_link'

Using an alternate port

If you'd like your Zulip server to use an HTTPS port other than 443, you can configure that as follows:

  1. Edit EXTERNAL_HOST in /etc/zulip/settings.py, which controls how the Zulip server reports its own URL, and restart the Zulip server with /home/zulip/deployments/current/scripts/restart-server.

  2. Add the following block to /etc/zulip/zulip.conf:

    [application_server]
    nginx_listen_port = 12345
    
  3. As root, run /home/zulip/deployments/current/scripts/zulip-puppet-apply. This will convert Zulip's main nginx configuration file to use your new port.

We also have documentation for a Zulip server using HTTP for use behind reverse proxies.

Customizing the outgoing HTTP proxy

To protect against SSRF, Zulip 4.8 and above default to routing all outgoing HTTP and HTTPS traffic through Smokescreen, an HTTP CONNECT proxy; this includes outgoing webhooks, website previews, and mobile push notifications. By default, the Camo image proxy will be automatically configured to use a custom outgoing proxy, but does not use Smokescreen by default because Camo includes similar logic to deny access to private subnets. You can override this default configuration if desired.

To use a custom outgoing proxy:

  1. Add the following block to /etc/zulip/zulip.conf, substituting in your proxy's hostname/IP and port:

    [http_proxy]
    host = 127.0.0.1
    port = 4750
    
  2. As root, run /home/zulip/deployments/current/scripts/zulip-puppet-apply. This will reconfigure and restart Zulip.

If you have a deployment with multiple frontend servers, or wish to install Smokescreen on a separate host, you can apply the zulip::profile::smokescreen Puppet class on that host, and follow the above steps, setting the [http_proxy] block to point to that host.

If you wish to disable the outgoing proxy entirely, follow the above steps, configuring an empty host value.

Optionally, you can also configure the Smokescreen ACL list. By default, Smokescreen denies access to all non-public IP addresses, including 127.0.0.1, but allows traffic to all public Internet hosts.

In Zulip 4.7 and older, to enable SSRF protection via Smokescreen, you will need to explicitly add the zulip::profile::smokescreen Puppet class, and configure the [http_proxy] block as above.

S3 file storage requests and outgoing proxies

By default, the S3 file storage backend bypasses the Smokescreen proxy, because when running on EC2 it may require metadata from the IMDS metadata endpoint, which resides on the internal IP address 169.254.169.254 and would thus be blocked by Smokescreen.

If your S3-compatible storage backend requires use of Smokescreen or some other proxy, you can override this default by setting S3_SKIP_PROXY = False in /etc/zulip/settings.py.

Putting the Zulip application behind a reverse proxy

Zulip is designed to support being run behind a reverse proxy server. This section contains notes on the configuration required with variable reverse proxy implementations.

Installer options

If your Zulip server will not be on the public Internet, we recommend, installing with the --self-signed-cert option (rather than the --certbot option), since Certbot requires the server to be on the public Internet.

Configuring Zulip to allow HTTP

Depending on your environment, you may want the reverse proxy to talk to the Zulip server over HTTP; this can be secure when the Zulip server is not directly exposed to the public Internet.

After installing the Zulip server as described above, you can configure Zulip to talk HTTP as follows:

  1. Add the following block to /etc/zulip/zulip.conf:

    [application_server]
    http_only = true
    
  2. As root, run /home/zulip/deployments/current/scripts/zulip-puppet-apply. This will convert Zulip's main nginx configuration file to allow HTTP instead of HTTPS.

  3. Finally, restart the Zulip server, using /home/zulip/deployments/current/scripts/restart-server.

Configuring Zulip to trust proxies

Before placing Zulip behind a reverse proxy, it needs to be configured to trust the client IP addresses that the proxy reports via the X-Forwarded-For header. This is important to have accurate IP addresses in server logs, as well as in notification emails which are sent to end users. Zulip doesn't default to trusting all X-Forwarded-For headers, because doing so would allow clients to spoof any IP address; we specify which IP addresses are the Zulip server's incoming proxies, so we know how much of the X-Forwarded-For header to trust.

  1. Determine the IP addresses of all reverse proxies you are setting up, as seen from the Zulip host. Depending on your network setup, these may not be the same as the public IP addresses of the reverse proxies. These can also be IP address ranges, as expressed in CIDR notation.

  2. Add the following block to /etc/zulip/zulip.conf.

    [loadbalancer]
    # Use the IP addresses you determined above, separated by commas.
    ips = 192.168.0.100
    
  3. Reconfigure Zulip with these settings. As root, run /home/zulip/deployments/current/scripts/zulip-puppet-apply. This will adjust Zulip's nginx configuration file to accept the X-Forwarded-For header when it is sent from one of the reverse proxy IPs.

  4. Finally, restart the Zulip server, using /home/zulip/deployments/current/scripts/restart-server.

nginx configuration

Below is a working example of a full nginx configuration. It assumes that your Zulip server sits at https://10.10.10.10:443; see above to switch to HTTP.

  1. Follow the instructions to configure Zulip to trust proxies.

  2. Configure the root nginx.conf file. We recommend using /etc/nginx/nginx.conf from your Zulip server for our recommended settings. E.g. if you don't set client_max_body_size, it won't be possible to upload large files to your Zulip server.

  3. Configure the nginx site-specific configuration (in /etc/nginx/sites-available) for the Zulip app. The following example is a good starting point:

    server {
            listen 80;
            listen [::]:80;
            location / {
                    return 301 https://$host$request_uri;
            }
    }
    
    server {
            listen                  443 ssl http2;
            listen                  [::]:443 ssl http2;
            server_name             zulip.example.com;
    
            ssl_certificate         /etc/letsencrypt/live/zulip.example.com/fullchain.pem;
            ssl_certificate_key     /etc/letsencrypt/live/zulip.example.com/privkey.pem;
    
            location / {
                    proxy_set_header        X-Forwarded-For $proxy_add_x_forwarded_for;
                    proxy_set_header        Host $http_host;
                    proxy_http_version      1.1;
                    proxy_buffering         off;
                    proxy_read_timeout      20m;
                    proxy_pass              https://10.10.10.10:443;
            }
    }
    

    Don't forget to update server_name, ssl_certificate, ssl_certificate_key and proxy_pass with the appropriate values for your deployment.

Apache2 configuration

Below is a working example of a full Apache2 configuration. It assumes that your Zulip server sits at https://internal.zulip.hostname:443. Note that if you wish to use SSL to connect to the Zulip server, Apache requires you use the hostname, not the IP address; see above to switch to HTTP.

  1. Follow the instructions to configure Zulip to trust proxies.

  2. Set USE_X_FORWARDED_HOST = True in /etc/zulip/settings.py and restart Zulip.

  3. Enable some required Apache modules:

    a2enmod ssl proxy proxy_http headers rewrite
    
  4. Create an Apache2 virtual host configuration file, similar to the following. Place it the appropriate path for your Apache2 installation and enable it (E.g. if you use Debian or Ubuntu, then place it in /etc/apache2/sites-available/zulip.example.com.conf and then run a2ensite zulip.example.com && systemctl reload apache2):

    <VirtualHost *:80>
        ServerName zulip.example.com
        RewriteEngine On
        RewriteRule ^ https://%{HTTP_HOST}%{REQUEST_URI} [R=301,L]
    </VirtualHost>
    
    <VirtualHost *:443>
      ServerName zulip.example.com
    
      RequestHeader set "X-Forwarded-Proto" expr=%{REQUEST_SCHEME}
    
      RewriteEngine On
      RewriteRule /(.*)           https://internal.zulip.hostname:443/$1 [P,L]
    
      <Location />
        Require all granted
        ProxyPass https://internal.zulip.hostname:443/ timeout=1200
      </Location>
    
      SSLEngine on
      SSLProxyEngine on
      SSLCertificateFile /etc/letsencrypt/live/zulip.example.com/fullchain.pem
      SSLCertificateKeyFile /etc/letsencrypt/live/zulip.example.com/privkey.pem
      # This file can be found in ~zulip/deployments/current/puppet/zulip/files/nginx/dhparam.pem
      SSLOpenSSLConfCmd DHParameters "/etc/nginx/dhparam.pem"
      SSLProtocol all -SSLv3 -TLSv1 -TLSv1.1
      SSLCipherSuite ECDHE-ECDSA-AES128-GCM-SHA256:ECDHE-RSA-AES128-GCM-SHA256:ECDHE-ECDSA-AES256-GCM-SHA384:ECDHE-RSA-AES256-GCM-SHA384:ECDHE-ECDSA-CHACHA20-POLY1305:ECDHE-RSA-CHACHA20-POLY1305:DHE-RSA-AES128-GCM-SHA256:DHE-RSA-AES256-GCM-SHA384
      SSLHonorCipherOrder off
      SSLSessionTickets off
      Header set Strict-Transport-Security "max-age=31536000"
    </VirtualHost>
    

    Don't forget to update ServerName, RewriteRule, ProxyPass, SSLCertificateFile, and SSLCertificateKeyFile as are appropriate for your deployment.

HAProxy configuration

Below is a working example of a HAProxy configuration. It assumes that your Zulip server sits at https://10.10.10.10:443see above to switch to HTTP.

  1. Follow the instructions to configure Zulip to trust proxies.

  2. Configure HAProxy. The below is a minimal frontend and backend configuration:

    frontend zulip
        mode http
        bind *:80
        bind *:443 ssl crt /etc/ssl/private/zulip-combined.crt
        http-request redirect scheme https code 301 unless { ssl_fc }
        default_backend zulip
    
    backend zulip
        mode http
        timeout server 20m
        server zulip 10.10.10.10:443 check ssl ca-file /etc/ssl/certs/ca-certificates.crt
    

    Don't forget to update bind *:443 ssl crt and server as is appropriate for your deployment.

Other proxies

If you're using another reverse proxy implementation, there are few things you need to be careful about when configuring it:

  1. Configure your reverse proxy (or proxies) to correctly maintain the X-Forwarded-For HTTP header, which is supposed to contain the series of IP addresses the request was forwarded through. Additionally, configure Zulip to respect the addresses sent by your reverse proxies. You can verify your work by looking at /var/log/zulip/server.log and checking it has the actual IP addresses of clients, not the IP address of the proxy server.

  2. Configure your proxy to pass along the Host: header as was sent from the client, not the internal hostname as seen by the proxy. If this is not possible, you can set USE_X_FORWARDED_HOST = True in /etc/zulip/settings.py, and pass the client's Host header to Zulip in an X-Forwarded-Host header.

  3. Ensure your proxy doesn't interfere with Zulip's use of long-polling for real-time push from the server to your users' browsers. This nginx code snippet does this.

    The key configuration options are, for the /json/events and /api/1/events endpoints:

    • proxy_read_timeout 1200;. It's critical that this be significantly above 60s, but the precise value isn't important.
    • proxy_buffering off. If you don't do this, your nginx proxy may return occasional 502 errors to clients using Zulip's events API.
  4. The other tricky failure mode we've seen with nginx reverse proxies is that they can load-balance between the IPv4 and IPv6 addresses for a given hostname. This can result in mysterious errors that can be quite difficult to debug. Be sure to declare your upstreams equivalent in a way that won't do load-balancing unexpectedly (e.g. pointing to a DNS name that you haven't configured with multiple IPs for your Zulip machine; sometimes this happens with IPv6 configuration).

PostgreSQL warm standby

Zulip's configuration allows for warm standby database replicas as a disaster recovery solution; see the linked PostgreSQL documentation for details on this type of deployment. Zulip's configuration builds on top of wal-g, our streaming database backup solution, and thus requires that it be configured for the primary and all secondary warm standby replicas.

In addition to having wal-g backups configured, warm standby replicas should configure the hostname of their primary replica, and username to use for replication, in /etc/zulip/zulip.conf:

[postgresql]
replication_user = replicator
replication_primary = hostname-of-primary.example.com

The postgres user on the replica will need to be able to authenticate as the replication_user user, which may require further configuration of pg_hba.conf and client certificates on the replica. If you are using password authentication, you can set a postgresql_replication_password secret in /etc/zulip/zulip-secrets.conf.

System and deployment configuration

The file /etc/zulip/zulip.conf is used to configure properties of the system and deployment; /etc/zulip/settings.py is used to configure the application itself. The zulip.conf sections and settings are described below.

When a setting refers to "set to true" or "set to false", the values true and false are canonical, but any of the following values will be considered "true", case-insensitively:

  • 1
  • y
  • t
  • yes
  • true
  • enable
  • enabled

Any other value (including the empty string) is considered false.

[machine]

puppet_classes

A comma-separated list of the Puppet classes to install on the server. The most common is zulip::profile::standalone, used for a stand-alone single-host deployment. Components of that include:

  • zulip::profile::app_frontend
  • zulip::profile::memcached
  • zulip::profile::postgresql
  • zulip::profile::redis
  • zulip::profile::rabbitmq

If you are using a Apache as a single-sign-on authenticator, you will need to add zulip::apache_sso to the list.

pgroonga

Set to true if enabling the multi-language PGroonga search extension.

[deployment]

deploy_options

Options passed by upgrade-zulip and upgrade-zulip-from-git into upgrade-zulip-stage-2. These might be any of:

  • --skip-puppet skips doing Puppet/apt upgrades. The user will need to run zulip-puppet-apply manually after the upgrade.
  • --skip-migrations skips running database migrations. The user will need to run ./manage.py migrate manually after the upgrade.
  • --skip-purge-old-deployments skips purging old deployments; without it, only deployments with the last two weeks are kept.

Generally installations will not want to set any of these options; the --skip-* options are primarily useful for reducing upgrade downtime for servers that are upgraded frequently by core Zulip developers.

git_repo_url

Default repository URL used when upgrading from a Git repository.

[application_server]

http_only

If set to true, configures Zulip to allow HTTP access; use if Zulip is deployed behind a reverse proxy that is handling SSL/TLS termination.

nginx_listen_port

Set to the port number if you prefer to listen on a port other than 443.

no_serve_uploads

To enable the the S3 uploads backend, one needs to both configure settings.py and set this to true to configure nginx. Remove this field to return to the local uploads backend (any non-empty value is currently equivalent to true).

queue_workers_multiprocess

By default, Zulip automatically detects whether the system has enough memory to run Zulip queue processors in the higher-throughput but more multiprocess mode (or to save 1.5GiB of RAM with the multithreaded mode). The calculation is based on whether the system has enough memory (currently 3.5GiB) to run a single-server Zulip installation in the multiprocess mode.

Set explicitly to true or false to override the automatic calculation. This override is useful both Docker systems (where the above algorithm might see the host's memory, not the container's) and/or when using remote servers for postgres, memcached, redis, and RabbitMQ.

rolling_restart

If set to a non-empty value, when using ./scripts/restart-server to restart Zulip, restart the uwsgi processes one-at-a-time, instead of all at once. This decreases the number of 502's served to clients, at the cost of slightly increased memory usage, and the possibility that different requests will be served by different versions of the code.

uwsgi_listen_backlog_limit

Override the default uwsgi backlog of 128 connections.

uwsgi_processes

Override the default uwsgi (Django) process count of 6 on hosts with more than 3.5GiB of RAM, 4 on hosts with less.

[postfix]

mailname

The hostname that Postfix should be configured to receive mail at.

[postgresql]

effective_io_concurrency

Override PostgreSQL's effective_io_concurrency setting.

listen_addresses

Override PostgreSQL's listen_addresses setting.

random_page_cost

Override PostgreSQL's random_page_cost setting

replication_primary

On the warm standby replicas, set to the hostname of the primary PostgreSQL server that streaming replication should be done from.

replication_user

On the warm standby replicas, set to the username that the host should authenticate to the primary PostgreSQL server as, for streaming replication. Authentication will be done based on the pg_hba.conf file; if you are using password authentication, you can set a postgresql_replication_password secret for authentication.

ssl_ca_file

Set to the path to the PEM-encoded certificate authority used to authenticate client connections.

ssl_cert_file

Set to the path to the PEM-encoded public certificate used to secure client connections.

ssl_key_file

Set to the path to the PEM-encoded private key used to secure client connections.

ssl_mode

The mode that should be used to verify the server certificate. The PostgreSQL default is prefer, which provides no security benefit; we strongly suggest setting this to require or better if you are using certificate authentication. See the PostgreSQL documentation for potential values.

version

The version of PostgreSQL that is in use. Do not set by hand; use the PostgreSQL upgrade tool.

[memcached]

memory

Override the number of megabytes of memory that memcached should be configured to consume; defaults to 1/8th of the total server memory.

[loadbalancer]

ips

Comma-separated list of IP addresses or netmasks of external load balancers whose X-Forwarded-For should be respected. These can be individual IP addresses, or CIDR IP address ranges.

[http_proxy]

host

The hostname or IP address of an outgoing HTTP CONNECT proxy. Defaults to localhost if unspecified.

port

The TCP port of the HTTP CONNECT proxy on the host specified above. Defaults to 4750 if unspecified.

listen_address

The IP address that Smokescreen should bind to and listen on. Defaults to 127.0.0.1.

enable_for_camo

Because Camo includes logic to deny access to private subnets, routing its requests through Smokescreen is generally not necessary. Set to true or false to override the default, which uses the proxy only if it is not the default of Smokescreen on a local host.