Hi Matt,
I can confirm that I now see bidirectional results for Lancaster.
Cheers,
John
On 09/07/2012 13:01, Matt Doidge wrote:
> Hello,
>> Leaving things over the weekend have cleared up a few issues - I now see
>> bidirectional throughput from all the sites I'm testing *except*
>> Lancaster. I still only see from Lancaster to Cambridge. This looks to
>> me like a possible firewall issue at Lancaster? I am using the default
>> PerfSonar ports.
>
> I think the problem is that the Perfsonar tests use random ports for
> their tests, and as I was trying to "be good" and keep the firewall
> running on our perfsonar boxes things just weren't working. However I
> dug up some documentation on how to collapse the wave function of these
> ports for the perfsonar services by adding a few lines to
> /etc/bwctld/bwctld.conf:
> peer_port 6001-6010
> iperf_port 5001-5010
> nuttcp_port 7001-7010
>
> & editing the "testports" line in /etc/owampd/owampd.conf (for boxes
> boxes doing latency checks).
>
> Then I could open the relevant ports on my local firewall (although in
> hindsight a wonder why I bother with so many holes in it!). It might
> have a few more holes then it stictly needs (I opened the ports listed
> on my Personar boxes' web interface as well as those mentioned in
> documentation) but hopefully tests will work normally again now that
> I've unrandomed things, opened even more ports & restarted the services.
>
> Cheers,
> Matt
>
>
>> For Sheffield, I don't see them in the GridPP community yet.
>> Cheers,
>> John
>>
>> On 06/07/2012 15:31, Matt Doidge wrote:
>>> Heya,
>>>
>>>> Just to be clear, as I nearly managed to confuse myself about this -
>>>> I'm
>>>> referring to the PerfSonar bandwidth tests (I think this thread started
>>>> by referring to IC's desire for iperf tests and got hijacked). It's
>>>> possible that my end is misconfigured - I've just disabled the latency
>>>> support on my box to make sure that isn't interfering (it was only
>>>> enabled in the first place to provide a quick sanity check), so
>>>> maybe we
>>>> should leave things alone until Monday?
>>>
>>> I too was referring to our Perfsonar boxes, further adding to the
>>> confusion and continuing the thread hijacking! I'm more then happy to
>>> leave things till Monday (I think my brain's already checked out for the
>>> weekend anyway).
>>>
>>> Have a good weekend all,
>>> Matt
>>>
>>>
>>>> Cheers,
>>>> John
>>>>
>>>> On 06/07/2012 15:17, Matt Doidge wrote:
>>>>> Heya,
>>>>>> I have results in both directions for all sites but Lancaster which
>>>>>> also
>>>>>> has single directionality in the latency tests.
>>>>>
>>>>> This implies a misconfiguration on Lancaster's iperf boxen. My
>>>>> instincts
>>>>> suggest that the machines' firewalls are getting in the way, but I
>>>>> have
>>>>> every port listed in the "local service" page for each machine open
>>>>> (to
>>>>> tcp connections anyway). Is there another port range we should open to
>>>>> receive tests? Or do we need to open ports to udp traffic too?
>>>>>
>>>>> Cheers,
>>>>> Matt
>>>>>
>>>>>>
>>>>>> cheers
>>>>>> alessandra
>>>>>>
>>>>>> On 06/07/2012 12:48, John Hill wrote:
>>>>>>> I'm a bit puzzled - I turned on the bandwidth test yesterday.
>>>>>>> Looking
>>>>>>> at the results today, I see numbers in both directions between
>>>>>>> Cambridge and Oxford, Birmingham and RAL. For Manchester (and
>>>>>>> QMUL) I
>>>>>>> only see throughput numbers from Cambridge, not in the reverse
>>>>>>> direction. For Lancaster, I only see results from Lancaster to
>>>>>>> Cambridge.
>>>>>>>
>>>>>>> This may of course be correct, but it isn't what I expected to see.
>>>>>>>
>>>>>>> John
>>>>>>>
>>>>>>> On 04/07/2012 11:48, Alessandra Forti wrote:
>>>>>>>> Hi,
>>>>>>>>
>>>>>>>> I've enabled the bandwidth, latency and traceroute tests against
>>>>>>>> all
>>>>>>>> sites listed in GridPP and turned off the services where
>>>>>>>> appropriate to
>>>>>>>> avoid confusion.
>>>>>>>>
>>>>>>>> No UK site seems to be part of LHCONE (?) and looking at the
>>>>>>>> number of
>>>>>>>> sites in ATLAS and Atlas I think ATLAS is the correct one.
>>>>>>>>
>>>>>>>> The basic seems configured. I only miss the dashboard
>>>>>>>>
>>>>>>>> https://perfsonar.usatlas.bnl.gov:8443/exda/?page=25&cloudName=UK
>>>>>>>>
>>>>>>>> which after a bit more memory effort I think is Jeremy's task.
>>>>>>>>
>>>>>>>> cheers
>>>>>>>> alessandra
>>>>>>>>
>>>>>>>> On 04/07/2012 09:30, Christopher J. Walker wrote:
>>>>>>>>> On 04/07/12 09:24, Alessandra Forti wrote:
>>>>>>>>>> ok so we stick to the generic names. I've seen other sites
>>>>>>>>>> haven't
>>>>>>>>>> bothered to distinguish them.
>>>>>>>>>>
>>>>>>>>>
>>>>>>>>> I've added Manchester and BHAM to QMUL's tests.
>>>>>>>>>
>>>>>>>>> Chris
>>>>>>>>>
>>>>>>>>>> On 04/07/2012 09:21, Christopher J. Walker wrote:
>>>>>>>>>>> On 04/07/12 09:13, Alessandra Forti wrote:
>>>>>>>>>>>> Do these machines have a problem with DNS aliases?
>>>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>>> They seem to report the real hostname.
>>>>>>>>>>>
>>>>>>>>>>> Chris
>>>>>>>>>>>
>>>>>>>>>>>>
>>>>>>>>>>>> On 03/07/2012 17:53, Alessandra Forti wrote:
>>>>>>>>>>>>> Hi,
>>>>>>>>>>>>>
>>>>>>>>>>>>> I've installed two machines in Manchester.
>>>>>>>>>>>>>
>>>>>>>>>>>>> sv220316.tier2.hep.manchester.ac.uk (latency)
>>>>>>>>>>>>> sv220317.tier2.hep.manchester.ac.uk (bandwidth)
>>>>>>>>>>>>>
>>>>>>>>>>>>> I added to the following communities
>>>>>>>>>>>>>
>>>>>>>>>>>>> GridPP
>>>>>>>>>>>>> ATLAS
>>>>>>>>>>>>> Atlas (don't know the difference with the previous one)
>>>>>>>>>>>>> LHCONE
>>>>>>>>>>>>>
>>>>>>>>>>>>> I think someone needs to add the machines here
>>>>>>>>>>>>>
>>>>>>>>>>>>> https://perfsonar.usatlas.bnl.gov:8443/exda/?page=25&cloudName=UK
>>>>>>>>>>>>>
>>>>>>>>>>>>>
>>>>>>>>>>>>>
>>>>>>>>>>>>>
>>>>>>>>>>>>> not sure about the full configuration yet though. I'm still
>>>>>>>>>>>>> looking at
>>>>>>>>>>>>> it.
>>>>>>>>>>>>>
>>>>>>>>>>>>> cheers
>>>>>>>>>>>>> alessandra
>>>>>>>>>>>>>
>>>>>>>>>>>>> On 25/06/2012 14:28, Jeremy Coles wrote:
>>>>>>>>>>>>>>> I'd also encourage everyone to get their perfsonar boxes up
>>>>>>>>>>>>>>> and
>>>>>>>>>>>>>>> running sooner rather than later for doing the routine
>>>>>>>>>>>>>>> ongoing
>>>>>>>>>>>>>>> testing.
>>>>>>>>>>>>>> I'd second that. In fact we do have a nominal plan for all
>>>>>>>>>>>>>> sites to
>>>>>>>>>>>>>> have perfsonar boxes installed by the end of July, and I'm
>>>>>>>>>>>>>> still
>>>>>>>>>>>>>> hoping that a few more sites will help us hit our (+4) June
>>>>>>>>>>>>>> target
>>>>>>>>>>>>>> (so far thanks to John at Cambridge we have one additional
>>>>>>>>>>>>>> site
>>>>>>>>>>>>>> (https://perfsonar.usatlas.bnl.gov:8443/exda/?page=25&cloudName=UK)
>>>>>>>>>>>>>>
>>>>>>>>>>>>>>
>>>>>>>>>>>>>>
>>>>>>>>>>>>>>
>>>>>>>>>>>>>>
>>>>>>>>>>>>>>
>>>>>>>>>>>>>> in addition to QMUL, Oxford and RAL.
>>>>>>>>>>>>>>
>>>>>>>>>>>>>> Jeremy
>>>>>>>>>>>>>>
>>>>>>>>>>>>>>
>>>>>>>>>>>>>>
>>>>>>>>>>>>>>
>>>>>>>>>>>>>>
>>>>>>>>>>>>>>
>>>>>>>>>>>>>>
>>>>>>>>>>>>>> On 25 Jun 2012, at 14:07, Ewan MacMahon wrote:
>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>> -----Original Message-----
>>>>>>>>>>>>>>>> From: Testbed Support for GridPP member institutes
>>>>>>>>>>>>>>>> [mailto:TB-
>>>>>>>>>>>>>>>> [log in to unmask]] On Behalf Of Simon Fayer
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>> We're looking to test our new network link here at Imperial
>>>>>>>>>>>>>>>> which
>>>>>>>>>>>>>>>> should
>>>>>>>>>>>>>>>> be capable of up to 40Gbit/s. Are there any 10+Gbit/s
>>>>>>>>>>>>>>>> connected
>>>>>>>>>>>>>>>> sites
>>>>>>>>>>>>>>>> which have iperf endpoints they'd be willing to let us test
>>>>>>>>>>>>>>>> against?
>>>>>>>>>>>>>>> Hi,
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>> As a temporary setup, everyone can feel free to use ours,
>>>>>>>>>>>>>>> it's on
>>>>>>>>>>>>>>> a 10G
>>>>>>>>>>>>>>> link but that's shared with the rest of Oxford; we've not
>>>>>>>>>>>>>>> actually
>>>>>>>>>>>>>>> got
>>>>>>>>>>>>>>> more than 5.5 out of it:
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>> t2-10gtest.physics.ox.ac.uk
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>> It's running iperf on the default port. I'd also encourage
>>>>>>>>>>>>>>> everyone to
>>>>>>>>>>>>>>> get their perfsonar boxes up and running sooner rather than
>>>>>>>>>>>>>>> later for
>>>>>>>>>>>>>>> doing the routine ongoing testing.
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>> Ewan
>>>>>>>>>>>>>
>>>>>>>>>>>>>
>>>>>>>>>>>>
>>>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>
>>>>>>>>
>>>>>>>>
>>>>>>
>>>>>>
>>>>
>>
|