[PROTON-1602] [Ruby] Possible memory leak in Ruby bindings?

• Type: Bug
• Status: Closed
• Priority: Critical
• Resolution: Fixed
• Affects Version/s: proton-c-0.17.0, proton-c-future
• Fix Version/s: proton-c-0.18.0
• Component/s: ruby-binding
• Environment: Centos 7
• Assignee: Alan Conway
• Reporter: Gregor Berginc
• Created: 29/Sep/17 06:13
• Updated: 24/Oct/17 08:44
• Resolved: 06/Oct/17 12:51

Description

While running integration tests we noticed that Ruby bindings utilise 100% CPU core and that memory consumption increases until the machine runs out of memory. The same Python app does not have this effect - low CPU utilisation.

I've used 0.17 version of qpid-proton-c from EPEL, as well as the current master. Ruby bindings in both cases built from the current master.

We are using the Container, i.e.

Qpid::Proton::Reactor::Container.new(...).run

This is very easily reproduced by the simple change in one of the tests:

diff --git a/proton-c/bindings/ruby/tests/test_container.rb b/proton-c/bindings/ruby/tests/test_container.rb
index d5b5c9a..455d63e 100644
--- a/proton-c/bindings/ruby/tests/test_container.rb
+++ b/proton-c/bindings/ruby/tests/test_container.rb
@@ -80,7 +80,8 @@ class ContainerSASLTest < Minitest::Test

     def on_connection_opened(event)
       super
-      event.container.stop
+      puts "keeping connection opened"
+#      event.container.stop
     end
   end

The change keeps the container opened. Looking at top, the CPU is at 100% and RAM constantly grows.

Is this intentional or by design? Is there a proper usage guide on how to access AMQP via Ruby?

Thanks!