HEPiX IPv6 working group meeting
Vidyo
HEPiX IPv6 working group meeting - Vidyo - Thursday 29 October 2015
Present: Marian Babik, Alastair Dewhurst,Thomas Finnern, Leslie Groer, Tiju Idiculla, Dave Kelsey, Francesco Prelz, Duncan Rand, Andrea Sciaba, Ulf Tigerstedt (joined late).
Apologies: Bruno Hoeft, Edoardo Martelli, Raja Nandakumar, Ulf Tigerstedt (for first part of meeting).
Introduction
DaveK welcomed Leslie Groer from Univ of Toronto. He has joined as they are interested in running IPv6-only worker nodes for use by ATLAS.
Minutes of September 2015 F2F meeting
No corrections needed. Accepted.
Actions and ongoing issues
- AndreaS. Dual-stack SAM3 monitoring.
- DaveK. Security Best Practices document.
- FrancescoP. Setup for XRootD testing.
Roundtable updates
a) ATLAS. AlastairD reported that there had been little activity until the last few days when they had received a request for Canada Tier 2 to run IPv6-only worker nodes. LeslieG explained that they would like to do this as early as Q2 of 2016. It was agreed that Univ of Toronto joining testing activites soon made sense with a test instance of the PANDA pilot job system at first (DuncanR has such a test system at Imperial). Alastair also reported that Squid3 now works with IPv6 and the Frontier package should work from version 3.5 onwards. In order to test this would need a dual-stack stratum 1 or Frontier launchpad. Alastair and Tiju agreed to test with dual-stack Squid.
b) AndreaS. No news from CMS. MarianB (who attended) reported on the ongoing action on dual-stack monitoring. They have been setting up a Nagios Srver for testing nodes in the testbed on a VM. They have IPv6-only to ensure traffic is IPv6. They will deploy the latest (new) version of SAM and Nagios. This work is already in an advanced state and should work. The new framework is a lot more flexible.
c) DuncanR. Imperial. Duncan reported that he has been working with Marian on the dual-stack perfSONAR mesh. This has been re-instated and is now working with some new sites added. Duncan has also been testing a DIRAC instance, trying to submit FTS jobs from the DIRAC command line. In perfSONAR, the IPv6 latency tests are currently disabled (and will be for a few months) because of performance limitations. The bi-directional traceroutes, however, are being run and archived. To see the traceroute history - access the host itself via the appropriate URL.
d) FrancescoP. INFN. Not much progress during the last month but he has prepared a container for xrootd redirection tests. He asked whether other sites could easily deploy Docker containers - silence suggested no! He has asked for an IPv6 training day at CNAF. He reports that the Google IPv6 stats are reporting latency problems with IPv6 over much of Europe.
e) ThomasF. DESY. Not much news but more IPv6 testing activity is just starting.
f) TijuI. RAL. Not much to report. The work to enable IPv6 routing at RAL is being planned but could now be next year.
g) UlfT. NDGF. They now have an IPv6-enabled SE but forgot to add AAAA record to DNS. When this was fixed they found an old bug - dcap client is IPv4-only. He is working on a fix for this. There is now a test version available that fixes the globus client IPv6 bug. This seems to make FTS3 work OK.
Testing plans
Ulf reports that he is in the process of re-instating Tony Wildish's IPv6 testbed. This should restart soon.
The ATLAS Canada Tier 2 tests mentioned above will start in the new year. Needs careful planning.
The group agrees that we should push for cvmfs to follow the standards based approach and favour IPv6 connections if they exist.
New Actions
Alastair/Tiju. Test Frontier with dual-stack Squid at RAL.
Dates of future meetings
Thursday 10th December, 2015. 16:00 - 17:00 CET (Vidyo)
F2F meeting on 21/22 Jan 2016 at CERN (lunchtime to lunchtime).
Notes by Dave Kelsey
10 December 2015