HEPiX IPv6 Working Group - Meeting using Vidyo
13 Mar 2014

Present:   Alastair Dewhurst, Marek Elias, Simon Fayer, Costin Grigoras, Bruno Hoeft, Dave Kelsey (Chair), Fernando Lopez Munoz, Kars Ohrenberg (but never successfully connected - Vidyo problems), Duncan Rand, Andrea Sciaba, Ulf Tigerstedt, Ramiro Voicu, Tony Wildish.

Apologies: Keith Chadwick, Thomas Finnern (joined, reported that he had no news and then left), Edoardo Martelli, Francesco Prelz (joined, reported that he had told the March pre-GDB meeting on batch systems about HTCondor and IPv6 and then left).

1. Dave welcomed all to the meeting. The agenda was agreed.

2. Roundtable reports.

a) CostinG reported that all publicly available ALICE services are now dual-stack but no IPv6 traffic as there are no IPv6 enabled clients yet. At a recent workshop, sites were made fully aware that they need to deploy IPv6 soon (this year or next year at latest).

b) AlastairD reported on IPv6 activities in ATLAS. He has spoken about IPv6 at several ATLAS meetings to raise its profile, including a recent software week. The initial goal is to focus on PANDA jobs (production analysis) working with an IPv6-enabled site. The development machines at CERN will be made dual-stack. DuncanR at Imperial is considering use of BIGPANDA.  Alastair has spoken to RAL, BNL and IN2P3 re possible testing. ATLAS is keen but the sites are not ready nor do they have imminent plans. There is an urgent need for development/test dual-stack systems at WLCG Tier 1s. ATLAS has started tests on Frontier and CVMFS. They have access to some student effort to test Squid3 on dual-stack.

Agreed: We should push Tier 1's for dual-stack test systems.

c) Fernando reported on PIC. They have installed dCache 2.8 in the testbed. GridFTP seems to work but srmcp fails - ls and rm work OK, but cp fails. A ticket has been opened on this. Ulf asks if they have multiple hostnames in the server certificate as this is known to cause problems. TonyW points out that PIC has an expired certificate (25 hours ago). Once this is fixed PIC would be ready to test with PhEDEx.

d) Imperial. Duncan reports he is having discussions with BIGPANDA and will be happy to test dual-stack. They have switched on dual-stack on the RESTful interface to the FTS3 server.

e) FZU. Marek reports that since the January F2F meeting they have been testing PerfSONAR on IPv6, integrating with DuncanR's mesh. They are still working on solving problems with SSL connections. HTTP works and Imperial say they may soon change to use of HTTP.

f) TonyW reports on the testbed status. INFN, DESY and FZU are OK. IHEP and PIC have expired host certificates. FNAL has some new problem. GridKa, CERN and NDGF all show "unable to connect/timed out".  Ulf says he sees transfers coming in to NDGF.   

ACTION: All sites should check if they see data transfers.

g) NDGF. Ulf reports that they are investigating pre-release versions of xrootd 4 and dCache. It is not working well. All commands are null-terminated. This is not in the specification, but if you don't null-terminate, nothing works.

h) CERN. Edoardo had sent an update by email. DHCPv6 opening for IT people's devices went very well; not a single problem of worse connectivity was reported. The IPv6 traffic didn't increase much though, because the CERN Windows PCs have a setting to prefer v4. It will be removed soon. Now they are confident to go with a full CERN deployment and only have to decide the date (probably April).

i) KIT. Bruno reported by email as he had audio problems. dCache 2.8 is installed and fts 3.0 is installed. Both are dual stack. The DNS Server is still pending.

j) DESY. Kars never successfully connected to Vidyo, but says he has no news.

3. Testing status and plans.

Tony W had reported the current data transfer testing status during the round-table reports (see above).

Future testing plans include:

a) Understand reason for GridFTP failures (TonyW's matrix).
b) Test dCache end points using PhEDEx.
c) Use cases to be tested by ATLAS.

4. Agenda items for the April 2014 F2F meeting at CERN.

Agreed that this should concentrate on plans for the 10th June pre-GDB workshop. This should include the agenda, what we cover, who talks, what tests need to be done before then?


Next meetings:
10-11 Apr 2014 F2F meeting at CERN.

Notes by Dave Kelsey.
9 Apr 2014