3.1.1 started breaking with getpid or cache_classes

[thebravoman]

On December 01 we've update to Spring 3.1.1

-    spring (3.1.0)
+    spring (3.1.1)

Since then one of two errors occur when we try to run the platform on the CI machine

It is this:

/usr/local/rvm/gems/ruby-2.6.5/gems/spring-3.1.1/lib/spring/server.rb:84:in `getpgid': No such process (Errno::ESRCH)
	from /usr/local/rvm/gems/ruby-2.6.5/gems/spring-3.1.1/lib/spring/server.rb:84:in `set_pgid'
	from /usr/local/rvm/gems/ruby-2.6.5/gems/spring-3.1.1/lib/spring/server.rb:39:in `boot'
	from /usr/local/rvm/gems/ruby-2.6.5/gems/spring-3.1.1/lib/spring/server.rb:14:in `boot'
	from /usr/local/rvm/gems/ruby-2.6.5/gems/spring-3.1.1/lib/spring/client/server.rb:10:in `call'
	from /usr/local/rvm/gems/ruby-2.6.5/gems/spring-3.1.1/lib/spring/client/command.rb:7:in `call'
	from /usr/local/rvm/gems/ruby-2.6.5/gems/spring-3.1.1/lib/spring/client.rb:30:in `run'
	from /usr/local/rvm/gems/ruby-2.6.5/gems/spring-3.1.1/bin/spring:49:in `<main>'

or this:

warning: parser/current is loading parser/ruby26, which recognizes
warning: 2.6.9-compliant syntax, but you are running 2.6.5.
warning: please see https://github.com/whitequark/parser#compatibility-with-ruby-mri.
/usr/local/rvm/gems/ruby-2.6.5/gems/spring-3.1.1/lib/spring/application.rb:101:in `block in preload': Spring reloads, and therefore needs the application to have reloading enabled.
Please, set config.cache_classes to false in config/environments/production.rb.
 (RuntimeError)
	from /usr/local/rvm/gems/ruby-2.6.5/gems/railties-6.1.4.1/lib/rails/initializable.rb:32:in `instance_exec'
	from /usr/local/rvm/gems/ruby-2.6.5/gems/railties-6.1.4.1/lib/rails/initializable.rb:32:in `run'
	from /usr/local/rvm/gems/ruby-2.6.5/gems/railties-6.1.4.1/lib/rails/initializable.rb:61:in `block in run_initializers'

This happen when we execute

RAILS_ENV=production rails s -d && kill -SIGTERM `cat tmp/pids/server.pid`

We need the last step of the process just to make sure the platform could start and we've not made a stupid configuration error in some of the production config files which has happened before

[thebravoman]

Downgraded to 3.1.0 - did not fix the issue

/usr/local/rvm/gems/ruby-2.6.5/gems/spring-3.1.0/lib/spring/server.rb:84:in `getpgid': No such process (Errno::ESRCH)
	from /usr/local/rvm/gems/ruby-2.6.5/gems/spring-3.1.0/lib/spring/server.rb:84:in `set_pgid'
	from /usr/local/rvm/gems/ruby-2.6.5/gems/spring-3.1.0/lib/spring/server.rb:39:in `boot'
	from /usr/local/rvm/gems/ruby-2.6.5/gems/spring-3.1.0/lib/spring/server.rb:14:in `boot'
	from /usr/local/rvm/gems/ruby-2.6.5/gems/spring-3.1.0/lib/spring/client/server.rb:10:in `call'
	from /usr/local/rvm/gems/ruby-2.6.5/gems/spring-3.1.0/lib/spring/client/command.rb:7:in `call'
	from /usr/local/rvm/gems/ruby-2.6.5/gems/spring-3.1.0/lib/spring/client.rb:30:in `run'
	from /usr/local/rvm/gems/ruby-2.6.5/gems/spring-3.1.0/bin/spring:49:in `<main>'

[rafaelfranca]

The cache_classes error is legitmate. Your need to fix it. Either you should disable cache classes, or disable spring in production. After you fix that, does the getpgid error still happens?

[thebravoman]

I guess it is a legitimate error. It surprised me that it started occurring about 2 weeks ago and it was not an issue for 3-4 years.

If I disable spring with DISABLE_SPRING=true then the error does not occur.

[rafaelfranca]

The getpgid is not legitimate. The cache classes started since 24cc80e

[rafaelfranca]

Does getpgid error still happens when you have cache_classes set to false in the test environment?

[thebravoman]

Haven't seen in happen. Only in production.

[rafaelfranca]

Why is spring being used in production?

[thebravoman]

Durring the years we've had cases where we make a change and an error In config/environments/production.rb

Tests and dev are passing and we only find out about the error when we deploy to production.

What we did then is as last step of the CI, before deploy, we start in production env as a way to check that there are no errors in the config.

We start and then kill the process

RAILS_ENV=production rails s -d && kill -SIGTERM `cat tmp/pids/

[rafaelfranca]

Ah, got it. Maybe you should disable spring in that command as well, or you will need to disable cache_classes.

DISABLE_SPRING=1 RAILS_ENV=production rails s -d && kill -SIGTERM `cat tmp/pids/

[thebravoman]

Yes this is exactly what I did In comment #664 (comment)

But I only found out about the env variable after creating the issue

[rafaelfranca]

Got it. I'm closing this issue since the fix is to either change cache_classes or disable spring. Thanks

[thebravoman]

Thanks. I think that's fair