Choose timezone
Your profile timezone:
Powered by Indico
v3.3.3-pre
The timings are approximate!
(meeting postponed from 26th October 2017)
Notes by Dave Kelsey.
Present: Martin Bly, Jiri Chudoba, Catalin Condurache, Alastair Dewhurst, Costin Grigoras, Kashif Hafez, David Kelsey, Edoardo Martelli, Raja Nandakumar, Kars Ohrenberg, Francesco Prelz, Andrea Sciaba.
Apologies: Tim Chown, Thomas Finnern, Bruno Hoeft, Duncan Rand.
Agenda: https://indico.cern.ch/event/679300/
Draft agenda was agreed.
Only outstanding action from the September F2F meeting was for Raja to provide an IPv6 summary table similar to that of ALICE. This action was completed by e-mail soon after the F2F meeting.
Italy. Francesco. He is currently preparing a tutorial for Italian colleagues for the Tuesday after this meeting BUT the INFN Tier1 has today been badly affected by a major burst water pipe. Francesco has also written a knowledge-base document on "Examples of RFC4890-compliant ICMP filter configuration".
See http://hepix-ipv6.web.cern.ch/content/examples-rfc4890-compliant-icmp-filter-configuration
Andrea. He has started sending out tickets to sites requesting status and plans. He is also working on a results table. So far, all UK sites have been sent tickets. France and Italy are next. There have been no requests for IPv6 support yet via GGUS.
Catalin. RAL services. The test version of FTS is now dual-stack and working for ATLAS. Next week they plan to move the production system to dual-stack and merge with the test system. RAL are now holding regular IPv6 coordination meetings.
Costin. ALICE. Together with Raja working on producing the same export of CEs and SEs. This work can be reviewed at the January F2F meeting. They would like to store and maintain a time-chart of the progress of move to dual-stack.
Edoardo. CERN. IPv6 VM status - still no fix from the router manufacturer for their ongoing problems. But they do now have a workaround which will be deployed soon. Edoardo will investigate why lx-plus has turned off IPv6.
Jiri. Prague. They have migrated from Torque to HTCondor and their new cluster will have all storage and worker nodes as dual-stack. They still have some problems with HTCondor and commands like lcg-cp using only IPv4. Debugging in real-time during the meeting proposed that setting enable-ipv6=true should be all that is needed for HTCondor.
Kars. DESY. The public login host is now dual-stack and work is ongoing on the CMS dCache instance.
Raja. LHCb. The statistics summary mentioned above in the Actions was a quick analysis using Python. He has discussed with the LHCb Dirac developers to aim for an xml-based web page with space for IPv6 status and comments. This should be available soon.
Ulf. NDGF. They are testing a new dCache version which is showing some strange behaviour. (cname, alias name has gone). A provided fix does not work.
Alastair. ATLAS. There have been tests of a NorduGrid site running IPv6-only WN in production and without the need for any NAT, this has successfully run all ATLAS jobs without problem. One problem with an event-index job which failed because of a message-broker at CERN being IPv4-only.
Nothing new
Andrea had already reported (see above) on his plans to send tickets to all sites.
Dave reported that he had been asked several times about a Tier2 site wishing to receive guidance on: Host-based firewalls using ip6tables, ICMP filtering advice for border router, guidance on AAAA records in DNS.
ip6tables. A brief discussion showed that in general many sites do not use this. All firewalling is done at the site border. Ulf does us IP6tables for dCache nodes.
ICMP filtering. See the knowledge base entry added by Francesco (details above).
- Conf call on the 7th December at 16:00 CET
- Face2face on the 25-27 January 2018 at CERN