EOFError: end of file reached issue with Net::HTTP

Question

I am using ruby-1.8.7-p302/Rails 2.3.11. I am trying to use FQL (Facebook API) to get stats for a link. Here's my code:

def stats(fb_post_url)
  url = BASE_URI + "?query=#{URI.encode("select like_count from link_stat where url=\"#{fb_post_url}\"")}"
  parsed_url = URI.parse(url)
  http = Net::HTTP.new(parsed_url.host, parsed_url.port)
  request = Net::HTTP::Get.new(parsed_url.request_uri)

  response = http.request(request)
  response.inspect
end

And here's the error:

EOFError: end of file reached
from /home/rahul/.rvm/rubies/ruby-1.8.7-p302/lib/ruby/1.8/net/protocol.rb:135:in `sysread'
from /home/rahul/.rvm/rubies/ruby-1.8.7-p302/lib/ruby/1.8/net/protocol.rb:135:in `rbuf_fill'
from /home/rahul/.rvm/rubies/ruby-1.8.7-p302/lib/ruby/1.8/timeout.rb:67:in `timeout'
from /home/rahul/.rvm/rubies/ruby-1.8.7-p302/lib/ruby/1.8/timeout.rb:101:in `timeout'
from /home/rahul/.rvm/rubies/ruby-1.8.7-p302/lib/ruby/1.8/net/protocol.rb:134:in `rbuf_fill'
from /home/rahul/.rvm/rubies/ruby-1.8.7-p302/lib/ruby/1.8/net/protocol.rb:116:in `readuntil'
from /home/rahul/.rvm/rubies/ruby-1.8.7-p302/lib/ruby/1.8/net/protocol.rb:126:in `readline'
from /home/rahul/.rvm/rubies/ruby-1.8.7-p302/lib/ruby/1.8/net/http.rb:2028:in `read_status_line'
from /home/rahul/.rvm/rubies/ruby-1.8.7-p302/lib/ruby/1.8/net/http.rb:2017:in `read_new'
from /home/rahul/.rvm/rubies/ruby-1.8.7-p302/lib/ruby/1.8/net/http.rb:1051:in `request'
from /home/rahul/.rvm/rubies/ruby-1.8.7-p302/lib/ruby/1.8/net/http.rb:1037:in `request'
from /home/rahul/.rvm/rubies/ruby-1.8.7-p302/lib/ruby/1.8/net/http.rb:543:in `start'
from /home/rahul/.rvm/rubies/ruby-1.8.7-p302/lib/ruby/1.8/net/http.rb:1035:in `request'
from /home/rahul/Work/Radr/lib/fb_stats.rb:13:in `stats'
from (irb):10

This seems to be happening only in case of the Facebook API. Also, I saw it suggested in some post this could be a bug in Net::HTTP.

Answer 1

If the URL is using https instead of http, you need to add the following line:

parsed_url = URI.parse(url)
http = Net::HTTP.new(parsed_url.host, parsed_url.port)
http.use_ssl = true

Note the additional http.use_ssl = true.

And the more appropriate code which would handle both http and https will be similar to the following one.

url = URI.parse(domain)
req = Net::HTTP::Post.new(url.request_uri)
req.set_form_data({'name'=>'Sur Max', 'email'=>'[e-mail]'})
http = Net::HTTP.new(url.host, url.port)
http.use_ssl = (url.scheme == "https")
response = http.request(req)

See more in my blog: EOFError: end of file reached issue when post a form with Net::HTTP.

Answer 2

In Ruby on Rails I used this code, and it works perfectly:

req_profilepic = ActiveSupport::JSON.decode(open(URI.encode("https://graph.facebook.com/me/?fields=picture&type=large&access_token=#{fb_access_token}")))

profilepic_url = req_profilepic['picture']

Answer 3

After doing some research, this was happening in Ruby's XMLRPC::Client library - which uses NET::HTTP. The client uses the start() method in NET::HTTP which keeps the connection open for future requests.

This happened precisely at 30 seconds after the last requests - so my assumption here is that the server it's hitting is closing requests after that time. I'm not sure what the default is for NET::HTTP to keep the request open - but I'm about to test with 60 seconds to see if that solves the issue.

Answer 4

I had the same problem, ruby-1.8.7-p357, and tried loads of things in vain...

I finally realised that it happens only on multiple calls using the same XMLRPC::Client instance!

So now I'm re-instantiating my client at each call and it just works:|

Answer 5

I find that I run into Net::HTTP and Net::FTP problems like this periodically, and when I do, surrounding the call with a timeout() makes all of those issues vanish. So where this will occasionally hang for 3 minutes or so and then raise an EOFError:

res = Net::HTTP.post_form(uri, args)

This always fixes it for me:

res = timeout(120) { Net::HTTP.post_form(uri, args) }

Answer 6

I had a similar problem with a request to a non-SSL service.

This blog suggested vaguely to try URI encoding the URL that is passed to 'get': http://www.loudthinking.org/2010/02/ruby-eoferror-end-of-file-reached.html

I took a shot at it, based on desperation, and in my limiting testing this seems to have fixed it for me. My new code is:

@http = Net::HTTP.new('domain.com')  
@http = @http.start    
url = 'http://domain.com/requested_url?blah=blah&etc=1'
req = Net::HTTP::Get.new(URI.encode(url))
req.basic_auth USERNAME, API_KEY
res = @http.request(req) 

Note that I use @http.start as I want to maintain the HTTP session over multiple requests. Other than that, you might like to try the most relevant part which is: URI.encode(url) inside the get call