Failed data upgrade to 2.9.0beta for multisite

Product Support

AI-generated summary

A user, hellekin, reports a failed data upgrade to Discourse 2.9.0beta on a multisite setup with two containers (web_only and data) on Ubuntu 20.04. The data container fails to bootstrap with a NoMethodError in the pups gem, specifically due to an undefined method 'each' for nil:NilClass error. Despite the failure, the user notes that the web_only container upgraded successfully, and running ./launcher start data allowed the existing container to start, showing the new version. JammyDodger confirms a similar upgrade on a test site succeeded, suggesting the issue might be resolved in newer builds. hellekin remains concerned about potential issues with future database migrations.

1

Although other hosts upgraded correctly, this multisite one with two containers (web_only and data) and no unofficial plugin has issues upgrading the data container:

./launcher rebuild data
x86_64 arch detected.
Ensuring launcher is up to date
Fetching origin
Launcher is up-to-date
Stopping old container
+ /usr/bin/docker stop -t 60 data
data
/usr/local/lib/ruby/gems/2.7.0/gems/pups-1.1.1/lib/pups.rb
/usr/local/bin/pups --stdin
I, [2022-03-21T14:08:36.336829 #1]  INFO -- : Reading from stdin
/usr/local/lib/ruby/gems/2.7.0/gems/pups-1.1.1/lib/pups/config.rb:70:in `combine_template_and_process_env': undefined method `each' for nil:NilClass (NoMethodError)
        from /usr/local/lib/ruby/gems/2.7.0/gems/pups-1.1.1/lib/pups/config.rb:17:in `initialize'
        from /usr/local/lib/ruby/gems/2.7.0/gems/pups-1.1.1/lib/pups/cli.rb:55:in `new'
        from /usr/local/lib/ruby/gems/2.7.0/gems/pups-1.1.1/lib/pups/cli.rb:55:in `run'
        from /usr/local/lib/ruby/gems/2.7.0/gems/pups-1.1.1/bin/pups:9:in `<top (required)>'
        from /usr/local/bin/pups:25:in `load'
        from /usr/local/bin/pups:25:in `<main>'
5bf317d1a66a44ec1d100a9da77210dfd7c10dd25f4cb999bd7c2ab69cd330dc
** FAILED TO BOOTSTRAP ** please scroll up and look for earlier error messages, there may be more than one.

I tried upgrading it several times, once after the ‘noop’ 2.9.0beta1 upgrade which was supposed to just work, and one today for beta2. Wondering…

The host is running Ubuntu 20.04.

2

FWIW, I’ve just rebuilt the data container on my test site, and that went through without a hitch. It is a new build from a few days ago though, so that may have helped.

(Digital Ocean, 1GB droplet, Ubunutu 20.04)

3

Since the web_only container upgraded successfully, and despite the data container failing to bootstrap, I ran ./launcher start data and the existing container started; reloading the admin page showed the new version instead.

I’m still curious about this bug since it may affect future upgrades that would require database migrations.