HEPiX IPv6 working group F2F meeting

Europe/Zurich
31/S-028 (CERN)

31/S-028

CERN

30
Show room on map
Description


WIFI access: pre-register your MAC address here (contact person: David Kelsey)

DRAFT agenda. Topics may change. Timings are approximate. 

The PIN for Vidyo has been distributed by email.

Participants
  • Alastair Dewhurst
  • Andrea Sciaba
  • Catalin Condurache
  • David Kelsey
  • Duncan Rand
  • Edoardo Martelli
  • Francesco Prelz
  • Kars Ohrenberg
  • Marian Babik
  • Martin Bly
  • qi mengyao
  • Ulf Bobson Severin Tigerstedt

 

HEPiX IPv6 Working Group Face to Face meeting

Held on Thursday 2/3 February 2017 at CERN

Attending:  Bruno Hoeft,  Alastair Dewhurst,  Andrea Sciaba, Catalin Condurache, David Kelsey, Duncan Rand, Edoardo Martelli, Francesco Prelz, Kars Ohrenberg, Marian Babik, Martin Bly, qi mengyao, Ulf Tigerstedt.

(Notes on Day 1 by Duncan Rand)

 

Agenda: https://indico.cern.ch/event/575321/

 

1. Review minutes and actions

Ulf suggested an additional entry to the agenda: IPv6 training at Manchester WLCG workshop.  The only previous action was to work on the CHEP paper.

 

2. Round table updates

LHCb. Raja announced that LHCb moved last service onto dual-stack - now ready for IPv6-only WN. Brunel WN is now running jobs. First job failed uploading to PIC - investigating.

ATLAS. Alastair Dewhurst  (see slides): Master JIRA ticket tracking rollout of IPv6 on services including Panda and Rucio. Would like sites to upgrade storage to dual-stack. Need to use CERN FTS service. BNL FTS server to be made dual-stack next week. Possible to see IPv6 transfers on monit.cern.ch. Jobs being tested at Brunel and QMUL. IPv6 WN temporarily switched off at QMUL. WN should be able to contact site squid which should be dual-stack. Should not be contacting T1 frontier’s directly.

CMS. Andrea Sciaba: Not much news. Andrea sent email to management suggesting an operational plan. Completely enable IPv6 on ITB and check works correctly. Secondly to contact T2s. Waiting for meeting to be scheduled. Edgar previously validated htcondor and glideinWMS. Farrukh moving on has delayed further progress. CMS xrootd redirector (at Bari?) and frontier service (at CERN) have been made dual-stack.

NDGF. Ulf: resurrected IPv6 only dCache so could be used for WLCG workshop training. ATLAS does a lot of production over IPv6.

CERN: no news.

Imperial. Duncan: Imperial Dirac service now able to run jobs on Brunel's IPv6-only WN.

IHEP (Beijing): qi mentioned he is interested in IPv6 address assignment. Dual-stack perfSONAR and test dCache operating. 

INFN. Francesco: Planning IPv6 transition at Milan. APNIC plot for Italy is a bit flat and now GARR (Italian NREN) is becoming interested. No news from CNAF.

Kars, DESY: business as usual. Trying out Cisco implementation of RFC6939.

RAL. Martin Bly: Making steps with central networking team to enable IPv6 on site core with a view to making April deadline. Border router dealing with IPv6 up to firewall (which does IPv6 in software). Next steps are to upgrade T1 core router to bring IPv6 to border of T1. Have an IPv6 addressing scheme for the site. Planning to have ability to make services dual-stack around start March. Plan to use xrootd proxy boxes to provide dual-stack access to Echo storage (not Castor). Is there any testing that can be done in Feb? Just connectivity testing. What about peering over LHCOPN? Should happen at beginning of March. perfSONAR will be used for testing.

KIT. Bruno: Not much news. Currently installing new storage system. Will be dual-stack. Estimated deployment around April. This will be the dual-stack production storage.

Nikhef. Dennis van Dok: Will try to include Sara for next report. At Nikef not much progress has been made. Large makeover of entire infrastructure. Should be able to go dual-stack storage quite soon.

3. Monitoring. Marian Babik

ETF (experiments test framework) IPv6 instance. Now tests CMS, ATLAS, LHCb and ‘IPv6 (ATLAS, CMS)’. etf.cern.ch now aggregates data from etf.cern.ch and etf-ipv6.cern.ch.

4. First look at CHEP papers

DK has put in an extension request for both papers to end of next week.

IPv6-only WN paper: Alastair went through the paper. Not very happy with software validation section and also VO sections, particularly CMS. Need to check authors list is correct. Should we include a plot of perfSONAR or FTS transfers? Send feedback either via github or directly to Alastair.

IPv6-only security paper: The meat of this paper is the check-list for developers and WLCG site system administrators and networking teams.

 

5. T0-T1-T2 status

The site status board does not have all the Tier-1s.

Action: Bruno to send tickets to Tier-1s to add themselves to the table.

Action: Edoardo to add columns recording whether dual-stack storage is ‘testbed’ or ‘production’ and what percentage of storage is available over IPv6 etc to twiki page.

Tier-2s: WLCG Ops see this as an operational issue and have volunteered to track it. Suggestion for sites represented around the table (e.g. DESY, IC) could add themselves to that table.

Dave mentioned email from WLCG Operations requesting that sites that have already deployed IPv6 make some notes describing the process. What should a site do to get going with IPv6? First step is perfsonar. Dual-stack storage is most useful, but probably the trickiest. In the past we have suggested the possibility that sites might be not be harshly treated for having poor reliability caused by implementing IPv6.

Action: Andrea to set up some sort of structure to produce some training material which might be usable at the WLCG workshop in Manchester, e.g.  a simple recipe as to what to do.

 

Day 2 - 3 Feb 2017 - notes by Ulf Tigerstedt

Meetings starts at 09:30

- Looking at agenda

CHEP papers - revisited

- Sysadmin checklist, 10 commandments:

   - one rfc referenced per commandment

- Papers should be ok with 2 week extension

- Should we do more graphs?

- Looking at the graphs we have, edoardo looks into getting

- Look into gather info from GOCDB and other sources, now that the BDIIs are winding down.

- Bruno LHCONE peering, data statistics.

- AP for project reps to gather stats with ipv6/ipv4 split

Next meetings

- next F2F meeting CERN, 15-16 May 2017

- Vidyo meetings

16 March 16:00-17:00 CET

12 April 16:00-17:00  CEST

 

 

 

 

There are minutes attached to this event. Show them.
  • Thursday, 2 February
    • 14:00 18:00
      Session 1 31/S-023

      31/S-023

      CERN

      22
      Show room on map
      • 14:00
        Introductions, agenda, note takers 10m
      • 14:10
        Review minutes and actions 10m
      • 14:20
      • 14:40
        Monitoring, ETF, etc. 25m
        Speaker: Marian Babik (CERN)
      • 15:05
        First look at status of CHEP papers 25m

        Latex files can be found:

        https://github.com/stfc/IPv6/tree/master/CHEP16

        We are entitled to 8 pages in total.

         

        Known issues:

        Software validation section is incomplete,  I am not sure what to put that is an update from previous proceedings.

        Not happy with the VO sections in particular CMS. 

        References need work.

         

         

         

         

      • 15:30
        Coffee 30m
      • 16:00
        Tier 0/1 status 45m
      • 16:45
        Tier 2 status 45m

        Current status?
        How do we determine/track the status?
        Guidance documentation - what? who? when?

      • 17:30
        Current technical issues 20m
  • Friday, 3 February
    • 09:00 13:00
      Session 2 31/S-028

      31/S-028

      CERN

      30
      Show room on map
      • 09:30
        Review agenda 5m
      • 09:35
        Monitoring the status of the migration to dual-stack 25m
      • 10:00
        SIgn-off on our two CHEP papers 30m
      • 10:30
        Coffee 30m
      • 11:00
        Other issues 1h

        LHCOPN/LHCONE
        PerfSONAR
        Data Transfer performance comparison

      • 12:00
        AOB and next meetings 15m
      • 12:15
        Review decisions and actions 15m