Passa al contenuto principale

Discourse

1. The scripts​

The official Discourse release uses its own Docker framework for installing everything that is needed for the application. Our scripts are just a shallow wrapper to this framework. They just customize a few settings to make it fit with the rest of docker-scripts apps (for example placing it behind revproxy).

We can get the scripts from https://gitlab.com/docker-scripts/discourse with ds pull:

ds pull discourse

cd /opt/docker-scripts/discourse/
tree

Because Discourse uses its own image and its own docker framework (which, by the way, is not docker compose), we don’t need a Dockerfile. We also override the commands build, create and config to make sure that they do nothing:

nano cmd/build.sh
nano cmd/create.sh
nano cmd/config.sh

nano misc/init.sh

misc/init.sh is called by the command ds init. It just clones the official discourse repo inside the directory.

To make the application, the config file settings.sh and the script ds.sh are used:

nano settings.sh
nano ds.sh

The file ds.sh is loaded automatically by the framework, when present, and can be used to override some framework functions, or to define new auxiliary functions. In this case, it is redefining cmd_start(), cmd_stop(), etc. in terms of the command launcher inside discourse_docker/, which is the official tool for managing the discourse containers. Two containers are managed: one for the app itself and one called mail-receiver.

cmd_make()

The most important function is cmd_make(), which does all the building and setup:

  • add a revproxy domain and get a ssl-cert
  • copy the app and mail-receiver config files
  • build the app and mail-receiver containers
  • create cron jobs, and make other configurations

Configuration files for the app and mail-receiver containers are copied from discourse_docker/samples/standalone.yml and discourse_docker/samples/mail-receiver.yml to discourse_docker/containers/ and then modified/customized.

note

They are copied only if they don't exist already. If they exist, they are not touched, because they can have manual customizations.

Some of the customizations that are done to discourse_docker/containers/talk.example.org.yml are these:

  • The exposed ports 80 and 443 are commented out, since the container will get the https requests from revproxy.

  • The container is connected to the docker-scripts network, using the docker args --network and --network-alias.

  • Volumes are mounted to the local directory /var/ds/talk.user1.fs.al/discourse_docker/ (instead of /var/disocurse/).

    note

    This is according to the docker-scripts policy of keeping all the configuration and data of an application in the same directory, which makes their management easier.

  • Email settings are assigned according to the values on settings.sh.

  • SSL is enabled (by uncommenting web.ssl.template.yml).

    note

    This also requires an SSL certificate. For this reason we are copying the LetsEncrypt certificate from revproxy to the appropriate place, where it can be loaded by the application as an external SSL certificate. Actually, a cron job is created that runs the command ds copy-ssl-cert periodically, each week.

  • We also add a line for each plugin listed on settings.sh, and customize some other settings.

The cron jobs that are created look like this:

  • /etc/cron.d/talk-user1-fs-al-restart

    # restart the application @/var/ds/talk.user1.fs.al each week
    0 0 * * 0  root  bash -l -c "ds @/var/ds/talk.user1.fs.al restart &> /dev/null"

  • /etc/cron.d/talk-user1-fs-al-copy-ssl-cert

    # copy the ssl cert @/var/ds/talk.user1.fs.al each week
    0 0 * * 0  root  bash -l -c "ds @/var/ds/talk.user1.fs.al copy-ssl-cert &> /dev/null"

2. Installation​

To install Discourse we need at least 4GB RAM free.

free -h
htop
glances
suggerimento

If there is less than 4GB RAM available, we can stop some applications. For example let's stop the incus container edu:

incus stop edu

Now let's initialize the container and build it:

ds init discourse @talk.user1.fs.al
cd /var/ds/talk.user1.fs.al/
nano settings.sh

Set ADMIN_EMAIL and remove some of the plugins. Then start building the container:

ds make

Note: This usually takes some time, especially with a small and not quite powerful VPS.

After it is done, let's check what happened:

ls
cd discourse_docker/

ls
ls containers/
nano containers/talk.user1.fs.al.yml
ls samples/
diff -u \
    samples/standalone.yml \
    containers/talk.user1.fs.al.yml \
    > d.diff
nano d.diff

nano containers/talk.user1.fs.al-mail-receiver.yml
diff -u \
    samples/mail-receiver.yml \
    containers/talk.user1.fs.al-mail-receiver.yml \
    > d.diff
nano d.diff
rm d.diff

docker ps | grep discourse
docker logs talk.user1.fs.al -f
docker logs talk.user1.fs.al-mail-receiver -f
note

Discourse usually does not need more than 1GB or 2GB of RAM, while running (only while installing it needs at least 4GB of RAM). So, we can start again the containers that we stopped before installing Discourse:

free -h

incus ls
incus start edu

3. Setup​

Open https://talk.user1.fs.al and continue the setup and configuration of the site.

The following sections explain some settings that are important, or specific to the way that we installed Discourse.

3.1 Sending emails​

The SMTP variables on settings.sh are used to set automatically the values of SMTP settings on the configuration file of the Discourse container, discourse_docker/containers/talk.user1.fs.al.yml, which look like this:

env:

  # . . . . . . . . . .

  ## TODO: The SMTP mail server used to validate new accounts and send notifications
  # SMTP ADDRESS, username, and password are required
  # WARNING the char '#' in SMTP password can cause problems!
  DISCOURSE_SMTP_ADDRESS: smtp.user1.fs.al
  DISCOURSE_SMTP_PORT: 25
  DISCOURSE_SMTP_USER_NAME:
  DISCOURSE_SMTP_PASSWORD:
  DISCOURSE_SMTP_ENABLE_START_TLS: true
  #DISCOURSE_SMTP_DOMAIN: discourse.example.com    # (required by some providers)
  #DISCOURSE_NOTIFICATION_EMAIL: noreply@discourse.example.com    # (address to send notifications from)
note

Despite the comment saying that username and password are required, we actually don’t need them because we are using the Simple SMTP Server, which allows applications from trusted hosts to send emails without authentication.

suggerimento

If you make some modifications to this yaml file, to apply them you have to rebuild the container:

cd /var/ds/talk.user1.fs.al/
cd discourse_docker/
./launcher rebuild talk.user1.fs.al

3.2 Discourse email settings​

To customize the discourse settings related to sending emails, we login as admin and go to "Admin / All Site Settings": https://talk.user1.fs.al/admin/site_settings/

There are too many settings there, but we can locate a setting quickly by typing its name on the filter box. Let's make sure that we set properly these settings:

For other email settings, that you might want to customize, look at https://talk.user1.fs.al/admin/site_settings/category/email

3.3 Test sending emails​

You can send a test email, from Discourse to your email address, at https://talk.user1.fs.al/admin/email/server-settings


To check the health of the mail server:

  1. Go to https://www.mail-tester.com/ and copy the displayed email address.

  2. Go to the email settings of Discourse and send a test email to the address given by the mail-tester.

  3. Go back to the mail-tester and check the score.

You should see something like this:


3.4 Mail receiver​

Discourse can send emails about the new topics and replies, notifications, etc. It would be nice if people can also reply or create new topics by email. This would make Discourse more similar to a mailing list. This documentation page explains how this is usually done: Configure incoming email with Mail-Receiver

It is using an additional container for receiving and processing incoming emails. The command ds make has already created the container talk.user1.fs.al-mail-receiver, which is based on the configuration containers/talk.user1.fs.al-mail-receiver.yml. This container is able to receive emails, process them accordingly, and push the necessary data to the container talk.user1.fs.al, using the API of discourse. However some additional steps are needed, in order to make it functional.

3.4.1 Add an MX record on the DNS​

On containers/talk.user1.fs.al-mail-receiver.yml we have set MAIL_DOMAIN to talk.user1.fs.al, so we need a DNS record like this:

talk.user1.fs.al.    IN    MX    10    smtp.user1.fs.al.

smtp.user1.fs.al is the name of the SMTP Server.

Without this DNS record, emails sent to an address @talk.user1.fs.al cannot be routed to our server.

cd /var/ds/nsd/
ls zones/
nano zones/user1.fs.al.db
ds restart
note

On zones/user1.fs.al.db, append these lines, and don't forget to update the serial number as well:

; Discourse
talk.user1.fs.al.    IN    A           188.245.242.143
talk.user1.fs.al.    IN    MX    10    smtp.user1.fs.al.

We can check that the record is already available with:

dig MX talk.user1.fs.al

3.4.2 Relay emails to mail-receiver​

The container talk.user1.fs.al-mail-receiver is hosted on the same server as smtp.user1.fs.al, so we cannot forward the SMTP ports to it, because they are being used for the mail server smtp.user1.fs.al. However, the simple SMTP server (smtp.user1.fs.al) can relay the mails for the domain talk.user1.fs.al. The section Relay domains describes in details how this can be done.

  1. Create a local command ds relay-setup, like this:

    cd /var/ds/smtp.user1.fs.al/
    mkdir -p cmd
    nano cmd/relay-setup.sh

    cmd/relay-setup.sh:

    cmd_relay-setup() {
        # create a config file for relay_domains
        cat <<EOF > config/relay_domains
    talk.user1.fs.al
    EOF

        # create a config file for transport_maps
        cat <<EOF > config/transport_maps
    talk.user1.fs.al    smtp:188.245.242.143:2501
    EOF

        # setup transport_maps
        ds exec postconf -e 'transport_maps=hash:/host/config/transport_maps'
        ds exec postmap /host/config/transport_maps

        # setup relay_domains
        ds exec postconf -e relay_domains=/host/config/relay_domains

        # reload postfix configuration
        ds exec postfix reload
    }

    It will set the settings relay_domains and transport_maps. Postfix will accept the emails of the domain @talk.user1.fs.al, and will transport (forward) them to the port 2501 of the server 188.245.242.143, using the smtp protocol.

    note

    • 188.245.242.143 is the public IP of the VPS itself

    • the port 2501 on the host is being forwarded to the docker container talk.user1.fs.al-mail-receiver. Check it with:

      docker ps | grep mail-receiver

  2. In order to automatically call the command ds relay-setup whenever the SMTP container is rebuilt, we can override the command ds config like this:

    cat <<'EOF' > cmd/config.sh
    rename_function cmd_config standard_config
    cmd_config() {
        standard_config
        ds relay-setup
    }
    EOF

    nano cmd/config.sh

    This creates a local config command, that extends the standard one.

  3. Enable the new configuration:

    ds make

    ls config/
    cat config/relay_domains
    cat config/transport_maps
    ds exec postconf relay_domains
    ds exec postconf transport_maps
note
Using a port different from 2501

If for some reason we want to use a port different from 2501, for example 2502, we can do it like this:

  1. Modify accordingly the script cmd/relay-setup.sh and call it: ds relay-setup

  2. Modify the configuration of mail-receiver and rebuild it:

    cd /var/ds/talk.user1.fs.al/discourse_docker/
    vim containers/talk.user1.fs.al-mail-receiver.yml
    ./launcher rebuild talk.user1.fs.al-mail-receiver

3.4.3 Set an API key​

The mails now should arrive to the mail-receiver, but it also needs an API key in order to push them to the Discourse container.

  1. Login as admin and go to: https://talk.user1.fs.al/admin/api/keys

  2. Create a new API key:

    • Description: mail-receiver container
    • User Level: Single User
    • User: system
    • Scope: Granular
    • Scopes: "email" / "receive emails"

  3. Copy the generated API key, and set it to the variable DISCOURSE_API_KEY at containers/talk.user1.fs.al-mail-receiver.yml

  4. Rebuild the container talk.user1.fs.al-mail-receiver:

    ./launcher rebuild talk.user1.fs.al-mail-receiver

3.4.4 Discourse configuration​

Now that email is being fed into Discourse, it’s time to explain to Discourse what to do with the emails it receives.

  1. Login as admin and navigate to "Admin / All Site Settings / Email": https://talk.user1.fs.al/admin/site_settings/category/email

  2. Change the following settings:

    • Check the setting "manual polling enabled".

    • In the "reply by email address" field, enter:

      replies+%{reply_key}@talk.user1.fs.al

    • Check the "reply by email enabled" setting.

    • Check the setting "email in".

  3. Other optional settings:

    • "email in allowed groups"

    • "unsubscribe via email footer"

You can use any address @talk.user1.fs.al as an address for category or group emails. This address has to be defined on the settings of the category or group.

3.4.5 Troubleshooting​

  1. From outside the server, send a test email to nobody@talk.user1.fs.al:

    apt install --yes swaks
    swaks --from somebody@gmail.com --to nobody@talk.user1.fs.al

    Alternatively, you can also send it from GMail or any other account.

  2. Check the logs on the SMTP server, to make sure that it arrived there:

    cd /var/ds/smtp.user1.fs.al/
    ds exec tail /var/log/mail.log -f

  3. Check the logs of mail-receiver to make sure that the message arrived there:

    docker logs talk.user1.fs.al-mail-receiver -f

    You should see a message like this:

    NOQUEUE: reject: RCPT from smtp.user1.fs.al[188.245.242.143]: 554 5.7.1 \
        <nobody@talk.user1.fs.al>: Recipient address rejected: \
        Mail to this address is not accepted. Check the address and try to send again?;

3.5 Install plugins​

On the configuration of the application (containers/talk.user1.fs.al.yml) there is a list of commands that clone plugin repos. Append the commands for the plugin(s) you want to install. Afterwards rebuild the container:

./launcher rebuild talk.user1.fs.al

To remove a plugin, remove it from the list and then rebuild the container.

suggerimento

It is a good idea to make a backup before installing a plugin.

4. Maintenance​

  • Backup and restore:

    cd /var/ds/talk.user1.fs.al/

    ds backup
    ds restore

    ds discourse

  • Upgrade:

    ds upgrade

  • Clone:

    ds clone talk-1.user1.fs.al
Cloning is usually needed to build a test site

Sometimes it is better to test some changes (for example installing a new plugin) on a clone of the site, and then apply them to the main site.

Cloning is done like this:

  1. Make a backup of the site.

  2. Initialize and build a new discourse instance for the clone.

  3. Restore the backup of the main site to the clone.

  4. Run discourse commands to rename the domain of the clone.

  5. Disable any outgoing emails on the clone (since it is used for testing).

note

Since cloning involves building a new Discourse instance, make sure that there are at least 4GB free RAM available (maybe by stopping some the containers).

  1. Let's edit /var/ds/_scripts/backup.sh and append these lines:

    cat <<EOF >> /var/ds/_scripts/backup.sh
    cd /var/ds/talk.user1.fs.al/
    ds backup
    find backup -type f -name "*.tar.gz" -mtime +5 -delete
    EOF

    nano /var/ds/_scripts/backup.sh

  2. Edit /var/ds/_scripts/update.sh and add these lines:

    # discourse
    cd /var/ds/talk.user1.fs.al/
    ds upgrade

5. More​

A useful admin guide is at: https://talk.user1.fs.al/t/admin-guide-getting-started/6 (you have to be logged in as admin).

Other docs/pages that might be interesting: