Environment Information

• JRuby version: jruby 9.2.11.1 (2.5.7) 2020-03-25 b1f55b1 OpenJDK 64-Bit Server VM 25.232-b09 on 1.8.0_232-b09 +jit [darwin-x86_64]
• OS: Darwin MBP.local 18.7.0 Darwin Kernel Version 18.7.0: Mon Feb 10 21:08:45 PST 2020; root:xnu-4903.278.28~1/RELEASE_X86_64 x86_64
• ActiveRecord *: 4.2.11.3

Expected Behavior

When I create an Enumerator over an ActiveRecord Relation using #to_enum, #peek on that enumerator does not always work. The exact behavior seems to vary a bit -- in my sample test, #peek throws StopIteration, but when I run this in larger environments, the Enumerator seems to hang. This may be tied to to AR Model size (and thus load time) -- I have a harder time reproducing with my small models, and easier with big, bloated objects.

https://gist.github.com/dgolombek/3123d5602ff6d76e96f2bac01b570210

Actual Behavior

Failures:

1. Enumerator with a single value in an AR array should not hang
   Failure/Error: raise "bad first element on first peek, #{iter.peek.id} != #{first.id}" unless iter.peek.id == first.id

   StopIteration:
   iteration reached an end

   uri:classloader:/jruby/kernel/enumerator.rb:25:in `next'

   uri:classloader:/jruby/kernel/enumerator.rb:34:in `peek'

   ./spec/simple_enum_spec.rb:4:in `check_iter'

   ./spec/simple_enum_spec.rb:25:in `block in '

   ./spec/spec_helper.rb:113:in `block in '

   ./spec/spec_helper.rb:112:in `block in '

   ./spec/spec_helper.rb:111:in `block in '

   ./spec/spec_helper.rb:100:in `block in '

   ./spec/spec_helper.rb:99:in `block in '

Using MRI 2.6.6, this works correctly. As mentioned above, when I deployed to prod, we saw uncontrolled thread growth, leading to server death, which I believe was tied to threads hanging in the enumerator (but could be tied to bugs with our handling of unexpected StopIteration, I'm still debugging that). I haven't gotten thread dumps yet, will work on doing so.