Add --single-ended flag to iperf2 and iperf3 #456

[vvidic]

Allows running throughput tests against servers running only
iperf2/iperf3 server on the default port.

[mfeit-internet2]

This needs to go through the process of incrementing the schema version for the test; failure to do that will result in confusing error messages when a task with the new switch is presented to a pScheduler that doesn't support it.

[vvidic]

On Thu, Mar 15, 2018 at 06:18:39PM +0000, Mark Feit wrote: This needs to go through the process of incrementing the schema version for the test; failure to do that will result in confusing error messages when a task with the new switch is presented to a pScheduler that doesn't support it.

Ok, but I don't know how to do that :) I also wanted to ask if we should have --single-ended-port=1234 for cases where iperf server is not using the default port?

…

-- Valentin Vidic Computer Systems Engineer - Expert Department of Computer Infrastructure and Services Croatian Academic and Research Network - CARNET Josipa Marohnica 5, HR-10000 Zagreb, Croatia tel: +385 1 6661 714, fax. +385 1 6661 635 gsm: +385 91 2480 919 www.CARNet.hr

[laeti-tia]

@mfeit-internet2 For the schema version increment, is it only the version field in https://github.com/perfsonar/pscheduler/blob/master/pscheduler-test-throughput/throughput/enumerate ? Shall we increase it to 1.1?

@vvidic Having an option like --single-ended-port=1234 would be nice, yes.

[mfeit-internet2]

@ToniN There's a bit more to it than that. If things are added or removed from what would be a valid test spec, there has to be a "schema": n member, where n is a version number (different from the package version) that gets incremented. The default is 1, which is where most of our stuff is.

I'll try and get something about how to do it written up and put in the pScheduler wiki this week.

[mfeit-internet2]

First cut of info on schema evolution: https://github.com/perfsonar/pscheduler/wiki/Schema-Evolution

[vvidic]

On Tue, Mar 27, 2018 at 03:20:32PM -0700, Mark Feit wrote: First cut of info on schema evolution: https://github.com/perfsonar/pscheduler/wiki/Schema-Evolution

Task fails with schema: 2, not sure what could be the problem? $ pscheduler task --debug --tool iperf2 throughput --dest xenial-amd64 --single-ended --single-ended-port 5555 2018-04-03T16:12:59 Debug signal ignored; already not debugging 2018-04-03T16:12:59 Debug discontinued 2018-04-03T16:12:59 Assistance is from localhost 2018-04-03T16:12:59 Forcing default slip of PT5M 2018-04-03T16:12:59 Converting to spec via https://localhost/pscheduler/tests/throughput/spec Submitting task... 2018-04-03T16:12:59 Fetching participant list 2018-04-03T16:12:59 Spec is: {"dest": "xenial-amd64", "single-ended": true, "single-ended-port": 5555, "schema": 2} 2018-04-03T16:12:59 Params are: {'spec': '{"dest": "xenial-amd64", "single-ended": true, "single-ended-port": 5555, "schema": 2}'} 2018-04-03T16:13:00 Got participants: {u'participants': [None], u'null-reason': u'No source specified'} 2018-04-03T16:13:00 Lead is None 2018-04-03T16:13:00 Null reason is No source specified 2018-04-03T16:13:00 Pinging https://jessie-amd64/pscheduler/ 2018-04-03T16:13:00 jessie-amd64 is up 2018-04-03T16:13:00 Posting task to https://jessie-amd64/pscheduler/tasks 2018-04-03T16:13:00 Data is {"test": {"type": "throughput", "spec": {"dest": "xenial-amd64", "single-ended": true, "single-ended-port": 5555, "schema": 2}}, "schema": 1, "tools": ["iperf2"], "schedule": {"slip": "PT5M"}} Unable to post task: Invalid test specification: Additional properties are not allowed (u'dest', u'schema', u'single-ended-port', u'single-ended' were unexpected)

…

-- Valentin Vidic Computer Systems Engineer - Expert Department of Computer Infrastructure and Services Croatian Academic and Research Network - CARNET Josipa Marohnica 5, HR-10000 Zagreb, Croatia tel: +385 1 6661 714, fax. +385 1 6661 635 gsm: +385 91 2480 919 www.CARNet.hr

[vvidic]

Schema now updated too, so please review again.

[vvidic]

Let me know if we need to update anything here?

[mfeit-internet2]

If you've tested the changes and are comfortable with it going to production, I'll pull it. My only hang-up was the schema numbering.

[vvidic]

On Thu, May 17, 2018 at 09:32:19AM -0700, Mark Feit wrote: If you've tested the changes and are comfortable with it going to production, I'll pull it. My only hang-up was the schema numbering.

Yes, I think it should be good. Maybe Antoine wants to give it a quick spin before the merge?

…

-- Valentin Vidic Computer Systems Engineer - Expert Department of Computer Infrastructure and Services Croatian Academic and Research Network - CARNET Josipa Marohnica 5, HR-10000 Zagreb, Croatia tel: +385 1 6661 714, fax. +385 1 6661 635 gsm: +385 91 2480 919 www.CARNet.hr

[laeti-tia]

If the single-ended-port is the port on the target iperf host, I think the default value should be the default iperf port (5201).

[laeti-tia]

Is there a reason to not use the regular 5201 iperf port?

[vvidic]

On Thu, May 17, 2018 at 10:10:25AM -0700, Antoine Delvaux wrote: If the single-ended-port is the port on the target iperf host, I think the default value should be the default iperf port (5201). > @@ -27,6 +27,8 @@ def test_all_options(self): "congestion": {"val": "reno" }, "client-cpu-affinity": {"val": "1" }, "server-cpu-affinity": {"val": "0" }, + "single-ended": {}, + "single-ended-port": { "val": "5000" }, Is there a reason to not use the regular 5201 iperf port?

I think this is just a test of some kind, so it should not matter much what you put in here. By default --single-ended uses the 5201 port and you only need --single-ended-port=X if it is not on this default port.

…

-- Valentin Vidic Computer Systems Engineer - Expert Department of Computer Infrastructure and Services Croatian Academic and Research Network - CARNET Josipa Marohnica 5, HR-10000 Zagreb, Croatia tel: +385 1 6661 714, fax. +385 1 6661 635 gsm: +385 91 2480 919 www.CARNet.hr

[laeti-tia]

All is good then, merged!