test-dgram-pingpong flaky on parallel runs

[jbergstroem]

It's been very frequent on the ppc machines during -J runs as of late: https://ci.nodejs.org/job/node-test-commit-plinux/nodes=ppcbe-fedora20/ws/test.tap/*view*/

[Trott]

I doubt this is the problem, but just in case: Is there any special firewalling or throttling on those machines? It looks like this test fires 1500 UDP packets in (let's say) half a second or so.

[jbergstroem]

@Trott: iptables looks squeaky clean and the vm admin dashboard doesn't point to any firewalling.

[Trott]

At least one other "flaky on PPC when run in parallel" test is a test that is also network intensive. In #4527, you report that test-http-regr-gh-2928 is flaky on PPC. That test checks that the HTTP flood prevention mechanism kicks in.

Maybe there's something the provider is doing that we can't see from the host itself? (I think all this stuff happens on localhost, so it wouldn't be on the subnet or anything like that, I imagine...)

[jbergstroem]

Regarding provider: that's why i checked the dashboard (there's usually ways to access firewalling from there). I'll create a support ticket and ask.

[Trott]

Maybe this test should just be moved to sequential. Maybe it's just too network (or other resrouce) intensive for some platforms.

[Trott]

Probable fix: #4615

[Trott]

I think #4615 fixed this. Please re-open if I'm mistaken.

[Trott]

Sad to be re-opening this: https://ci.nodejs.org/job/node-test-binary-arm/934/RUN_SUBSET=0,nodes=pi2-raspbian-wheezy/tapTestReport/test.tap-138/