Magic link emails fail in production: “Missing host to link to” in MagicLinkMailer#sign_in_instructions

[KingSolomun79]

Use this sanitized version. I replaced the real domain/email with placeholders.

Title

Magic link emails fail in production: “Missing host to link to” in MagicLinkMailer#sign_in_instructions

Discussion text

Hi, I’m self-hosting Fizzy and ran into an issue where login emails are not being sent.

The app itself seems to boot normally. Puma starts successfully, the /up health check returns 200 OK, and Solid Queue workers start correctly.

When I submit the login form, the request reaches the backend correctly:

Started POST "/session"
Processing by SessionsController#create as TURBO_STREAM
Parameters: {"email_address" => "user@example.com"}
Enqueued ActionMailer::MailDeliveryJob ... "MagicLinkMailer", "sign_in_instructions", "deliver_now"
Redirected to https://example.com/session/magic_link
Completed 302 Found

So the login form is working, and the mail delivery job is being created.

The problem happens when Solid Queue performs the mailer job:

Performing ActionMailer::MailDeliveryJob ...
Rendered layout layouts/mailer.html.erb
Error performing ActionMailer::MailDeliveryJob ... ActionView::Template::Error
(Missing host to link to! Please provide the :host parameter, set default_url_options[:host], or set :only_path to true)

The stack trace points here:

app/views/mailers/magic_link_mailer/sign_in_instructions.html.erb:15
app/mailers/magic_link_mailer.rb:6:in 'MagicLinkMailer#sign_in_instructions'
app/jobs/concerns/account_tenanted.rb:36:in 'AccountTenanted#with_account_context'

This makes it look like the email fails while rendering the magic-link URL, before it reaches SMTP.

My deployment has environment variables similar to:

SERVICE_FQDN_FIZZY=example.com
SERVICE_URL_FIZZY=https://example.com
MAILER_FROM_ADDRESS=no-reply@example.com

SMTP_ADDRESS=smtp.example.com
SMTP_AUTHENTICATION=login
SMTP_DOMAIN=example.com
SMTP_ENABLE_STARTTLS_AUTO=true
SMTP_PORT=465
SMTP_USERNAME=no-reply@example.com
SMTP_TLS=true

I also tried adding:

APP_HOST=example.com
APP_URL=https://example.com

But the same error still happens, so maybe Fizzy does not currently read those env vars into config.action_mailer.default_url_options or Rails.application.routes.default_url_options.

My guess is that production needs something like this somewhere:

config.action_mailer.default_url_options = {
  host: ENV.fetch("APP_HOST"),
  protocol: "https"
}

Rails.application.routes.default_url_options = {
  host: ENV.fetch("APP_HOST"),
  protocol: "https"
}

What is the intended way to configure the public app host/domain for magic-link emails in a self-hosted deployment? Is there a specific environment variable Fizzy expects for this, or should this be added to the production config?

[KingSolomun79]

The issue was not SMTP. The mailer job was failing before delivery because Rails could not generate the absolute URL inside the magic-link email.

The error was:

ActionView::Template::Error
Missing host to link to! Please provide the :host parameter, set default_url_options[:host], or set :only_path to true

The fix was to add:

BASE_URL=https://example.com

Then restart the app/container.

After adding BASE_URL, the mailer job completed successfully:

Rendered layout layouts/mailer.html.erb
Rendered layout layouts/mailer.text.erb
Performed ActionMailer::MailDeliveryJob

In my case, the hosting platform automatically provided variables like SERVICE_URL_* / SERVICE_FQDN_*, but Fizzy expects BASE_URL for Rails URL generation in production.

So for self-hosted deployments, make sure BASE_URL is set to the public HTTPS URL of the Fizzy instance.