perfSONAR User Q&A and Other Discussion

[Andrew Lake <>]

Hi Garrett,

Just to clarify, I don’t think what you said is entirely true for all tests. From your previous communications I think you were running just ping tests which do indeed exhibit this incorrect behavior (something Mark and I have discussed and acknowledge need to be fixed). The throughput and latency tests do not behave in this way and prefer IPV6 without any options if both ends have IPv6 available. Are you seeing the same behavior in these tests?

Thanks,

Andy

On October 28, 2019 at 4:07:40 PM, Garret Peirce () wrote:

Hi all,

I ran into an issue with PS and IPv6 recently I'd heard these lists had an interest in.

I'm not currently subscribed to either list here appreciate if any responses could include me directly.

I've an IPv6 only PS host that won't properly resolve the remote host and therefore it's test fails.

I can resolve the remote host and ping6 it without issue.

After inquiring with Mark Feit from I2, it appears this is a known PS node behavior. Perfsonar uses v4 unless the v6 switch is included. He is going to initiate a ticket to review the behavior towards implementing a patch.  The current workaround requires adding '--ip-version 6' to the test command.

Participating in a group of nodes sharing a test JSON file, including the option is a bit painful as I'll have to edit the file for v6 only use after any changes - which I've yet to get back to try.  Once seems painful enough when there seems to be an easier solution, i.e. having perfsonar follow RFC6724 for address selection.  Perhaps there are reasons to prefer manual/fixed selection, so manual options should probably remain.  We're later than many to the game, so I was surprised to find this little chestnut, but sounds like we'll see a PS patch to improve its current behavior.

-- 

Garry Peirce
Network Architect
Networkmaine, University of Maine System US:IT
207-581-3539

--
To unsubscribe from this list: https://lists.internet2.edu/sympa/signoff/perfsonar-user