Comment 17 for bug 1224001

At this stage I feel enough confident to say that the gate is failing because rpc_loop or sync_routers_task hang on subprocess.popen.communicate
This happens especially when executing arping. The actual reason is unknown.
One possibility is that arping is returning a very large output (unlikely), whereas another is that arping itself does not complete.
subprocess.popen.communicate will not stop in this case causing the hang.

I suggest to do the same mitigation openstack.common for execute() and tweaking the configuration option for not executing arping on the gate.