IOError when respond with large body

[li-thy-um]

This will cause the server process abort.

I, [2018-02-28T08:09:48.730775 #1]  INFO -- : Midori 0.5.2 is now running on 0.0.0.0:3000
I, [2018-02-28T08:09:50.043817 #1]  INFO -- : 172.25.0.1 - - "GET /documents HTTP/1.0" 200 OK 0.034884
I, [2018-02-28T08:09:50.360293 #1]  INFO -- : 172.25.0.1 - - "GET /origin_websites HTTP/1.0" 200 OK 0.286312
I, [2018-02-28T08:09:50.437701 #1]  INFO -- : 172.25.0.1 - - "GET /output_columns HTTP/1.0" 200 OK 0.076205
I, [2018-02-28T08:09:51.133797 #1]  INFO -- : 172.25.0.1 - - "GET /compose_organizations HTTP/1.0" 200 OK 0.693688
/app/vendor/bundle/ruby/2.5.0/gems/murasaki-0.2.0/lib/murasaki/event_loop.rb:62:in `fileno': closed stream (IOError)
    from /app/vendor/bundle/ruby/2.5.0/gems/murasaki-0.2.0/lib/murasaki/event_loop.rb:62:in `deregister'
    from /app/vendor/bundle/ruby/2.5.0/gems/midori.rb-0.5.2/lib/midori/connection.rb:63:in `close_connection'
    from /app/vendor/bundle/ruby/2.5.0/gems/midori.rb-0.5.2/lib/midori/connection.rb:33:in `block in listen'
    from /app/vendor/bundle/ruby/2.5.0/gems/murasaki-0.2.0/lib/murasaki/event_loop.rb:75:in `block in run_once'
    from /app/vendor/bundle/ruby/2.5.0/gems/murasaki-0.2.0/lib/murasaki/event_loop.rb:74:in `select'
    from /app/vendor/bundle/ruby/2.5.0/gems/murasaki-0.2.0/lib/murasaki/event_loop.rb:74:in `run_once'
    from /app/vendor/bundle/ruby/2.5.0/gems/murasaki-0.2.0/lib/murasaki/event_loop.rb:101:in `start'
    from /app/vendor/bundle/ruby/2.5.0/gems/midori.rb-0.5.2/lib/midori/runner.rb:44:in `start'
    from /app/midori.rb:6:in `<main>'

[dsh0416]

Bug confirmed. Can you provide any sample code?

[li-thy-um]

It's hard to provide sample code, but I'm trying.

I removed the Nginx reverse proxy in between, and did the test again. It will response the data successfully, but the server will shutdown.

get '/multi_aggs' do
   # here it will do elasticsearch aggregation search
   MultiAggs::Index.new(request.query_params, user).to_json
end

RestClient::Request.execute(method: :get, url: url,
      timeout: 5,
      headers: {params: params, Authorization: @auth_token})

I, [2018-03-01T04:48:23.105229 #1]  INFO -- : Midori 0.5.2 is now running on 0.0.0.0:3000
287465 # This is the response body's size of request bellow
I, [2018-03-01T04:48:38.949328 #1]  INFO -- : 172.25.0.1 - - "GET /multi_aggs HTTP/1.1" 200 OK 0.971632
320426
I, [2018-03-01T04:48:47.213653 #1]  INFO -- : 172.25.0.1 - - "GET /multi_aggs HTTP/1.1" 200 OK 1.264781
361135
I, [2018-03-01T04:48:56.695555 #1]  INFO -- : 172.25.0.1 - - "GET /multi_aggs HTTP/1.1" 200 OK 1.449826
286526
I, [2018-03-01T04:49:06.100274 #1]  INFO -- : 172.25.0.1 - - "GET /multi_aggs HTTP/1.1" 200 OK 1.100490
180645
I, [2018-03-01T04:49:11.421074 #1]  INFO -- : 172.25.0.1 - - "GET /multi_aggs HTTP/1.1" 200 OK 0.465922
/app/vendor/bundle/ruby/2.5.0/gems/murasaki-0.2.0/lib/murasaki/event_loop.rb:62:in `fileno': closed stream (IOError)
    from /app/vendor/bundle/ruby/2.5.0/gems/murasaki-0.2.0/lib/murasaki/event_loop.rb:62:in `deregister'
    from /app/vendor/bundle/ruby/2.5.0/gems/midori.rb-0.5.2/lib/midori/connection.rb:63:in `close_connection'
    from /app/vendor/bundle/ruby/2.5.0/gems/midori.rb-0.5.2/lib/midori/connection.rb:33:in `block in listen'
    from /app/vendor/bundle/ruby/2.5.0/gems/murasaki-0.2.0/lib/murasaki/event_loop.rb:75:in `block in run_once'
    from /app/vendor/bundle/ruby/2.5.0/gems/murasaki-0.2.0/lib/murasaki/event_loop.rb:74:in `select'
    from /app/vendor/bundle/ruby/2.5.0/gems/murasaki-0.2.0/lib/murasaki/event_loop.rb:74:in `run_once'
    from /app/vendor/bundle/ruby/2.5.0/gems/murasaki-0.2.0/lib/murasaki/event_loop.rb:101:in `start'
    from /app/vendor/bundle/ruby/2.5.0/gems/midori.rb-0.5.2/lib/midori/runner.rb:44:in `start'
    from /app/midori.rb:6:in `<main>'

[li-thy-um]

Gems:

gem 'activesupport'
gem 'tzinfo-data'
gem 'midori.rb'
gem 'elasticsearch'
gem 'monadic'
gem 'jbuilder'
gem 'nokogiri'
gem 'redis'

[dsh0416]

This seems to be the problem that midori accidentally closed the stream multiple times. A patch would be applied soon.

[luikore]

Any update on this bug?

[dsh0416]

@luikore There has been a temp fix merged, which has already fixed this problem. But it may take some time to rewrite the part with the fd number to fix the problem more elegantly.

[dsh0416]

Fixed in the v0.5.4 with the update of murasaki.

Thank you guys so much for reporting and tracking this issue. Sorry for the late fix.