Description of problem:
Each Puppet runs during the deploy are taking around 15-20 minutes for large and busy env with more than 1200 containers per controller

Version-Release number of selected component (if applicable):
Red Hat OpenStack Platform 13

How reproducible:
Always

Steps to Reproduce:
1. Deploy an overcloud and create more private networks with almost 1200 containers / controller.
2. Run a deploy command or ansible playbook to deploy / upgrade env.
3. 

Actual results:
Puppet run is slower

Expected results:
Puppet runs should be faster and not loop in reading /proc/mounts multiple times.


Additional info:
Just controllers alone took 16 hours to upgrade

Upon further checking puppet is kind of looping in reading /proc/mounts file which is really slow. 

Found the code responsible for slowness in
/usr/share/ruby/vendor_ruby/puppet/util/selinux.rb

~~~~~~~~~~~~~~~~
  def read_mounts
    mounts = ""
    begin
      if File.method_defined? "read_nonblock"
        # If possible we use read_nonblock in a loop rather than read to work-
        # a linux kernel bug.  See ticket #1963 for details.
        mountfh = File.open("/proc/mounts")
        mounts += mountfh.read_nonblock(1024) while true   << Here
      else
        # Otherwise we shell out and let cat do it for us
        mountfh = IO.popen("/bin/cat /proc/mounts")
        mounts = mountfh.read        << Here
      end
    rescue EOFError
      # that's expected
    rescue
      return nil
    ensure
      mountfh.close if mountfh
    end
~~~~~~~~~~~~~~~~~~

After commenting both reads, the puppet run took 47 sec in busy env, non busy env is taking around 9s.