Issue with HA console when ruby-debug is loaded

[dwbutler]

I tried out the console in HA mode following the instructions here:

https://github.com/andreasronge/neo4j/wiki/Neo4j%3A%3Aha-cluster

It didn't work for me at first - the console kept opening the same database as the rails server and going into readonly mode. Eventually I tracked the problem down to this line:

https://github.com/andreasronge/neo4j/blob/master/lib/neo4j/rails/ha_console/railtie.rb#L12

It turns out in my app, I load the 'ruby-debug' gem. This is loaded by Bundler This in turn loads IRB, which causes both the rails console and server to have IRB loaded.

I worked around the issue by doing the following:

Gemfile:

group :development do
  gem 'ruby-debug', :require => false
end

config/application.rb:

require 'neo4j/rails/ha_console/railtie' if Rails.env.development?
...
module MyRailsApp
  class Application < Rails::Application
    ...
    require 'ruby-debug' if Rails.env.development?

I'm not sure if there's a better way to do this. It's difficult to tell whether one is running in the Rails server or in the console.

[andreasronge]

Aha, good to know. No I don't if there is a better way.

[ruprict]

I don't think this will work for Pry, as it is defined everywhere in development. As such, both server and console get 2 as a server_id.

Our fix was to put the contents of the railtie into config/environments/development.rb, check for Rails::Console:

 server_id = ((defined? Rails::Console)) ? 2 : 1
 config.neo4j['enable_ha'] = true
 config.neo4j['ha.server_id'] = server_id
 config.neo4j['ha.server'] = "localhost:600#{server_id}"
 config.neo4j['ha.pull_interval'] = '500ms'
 config.neo4j['ha.discovery.enabled'] = false
 config.neo4j['ha.initial_hosts'] = [1,2,3].map{|id| "localhost:500#{id}"}.join(',')
 config.neo4j['ha.cluster_server'] = "localhost:500#{server_id}"
 config.neo4j.storage_path = File.expand_path("db/ha_neo_#{server_id}", Object::Rails.root)
 puts "Config HA cluster, ha.server_id: #{config.neo4j['ha.server_id']}, db: #{config.neo4j.storage_path}"

Make sure you comment out the ha_console/railtie line in application.rb if you go this route. We are still testing, but it's working in our first glance.

[dwbutler]

Cool! This fix works for me.

[ruprict]

BTW, if you want the neo-shell back, you can add

config.neo4j['enable_remote_shell'] = "port=133#{server_id}" 

To this config, then from a terminal prompt neo4j-shell -port 1331 (or whatever the port you want is) you have the shell back.

NOTE: The neo4j-shell command MUST be run from a prompt where JRUBY is not the current ruby. So, I am running RVM, I run it from a prompt where I am using ruby 2.0 and it works. If you try to run the neo4j-shell from a prompt where the ruby is JRuby, you'll get a Java error about JLine something or other (there is an issue on this)

Anyway, I was glad to have the shell back.

[dwbutler]

@ruprict, do you plan to submit a pull request for this?

[ruprict]

I could, but have to figure out what will be in that PR. The fixes I have here are pretty much doc changes. From what I have found, the things we need to be defined (Rails::Console and Pry) are not when the railtie is called. As such, we have to take the code out of the railtie and put it in the development.rb config file.

I guess we could put this in an initializer (complete with a if Rails.env.development?), but that seems like a lateral move, at best.

I'd love to hear other opinions and/or corrections, as I am prepared to be wrong about this.

[dwbutler]

I dug through the Rails source a bit and I think I found the solution. It appears that you can add block to a railtie that will be run when the console is loaded. Something like this should work

class Neo4j::Railtie < Rails::Railtie
  config.neo4j['ha.server_id'] = 1

  console do
    config.neo4j['ha.server_id'] = 2
  end

  after_initialize do
    config.neo4j['ha.server'] = "localhost:600#{config.neo4j['ha.server_id']}"
    # etc
  end
end

[ruprict]

Nice, that's great. Have you tried this? If so, you should definitely PR it.

Good work.

[dwbutler]

I've not actually tried it, but I'll try it soon when I find some time.

[thekendalmiller]

#261 to fix this

[dwbutler]

Thanks @thekendalmiller!