Choose timezone
Your profile timezone:
HEPiX IPv6 working group F2F meeting
Description
DRAFT agenda. Topics may change. Timings are approximate.
Please REGISTER (by 14th Sep) if you plan to attend the meeting in person at CERN.
HEPiX IPv6 F2F Meeting notes 20180918 - notes by Duncan Rand
Present: Edoardo, Duncan, Andrea, Costin, Kars, Francesco, Bruno, David
Remote: Ulf, Tim, Kashif, Michael, Fernando
Agenda: https://indico.cern.ch/event/754800/
Review minutes and actions and other urgent topics.
Submit abstract to HEPiX in Barcelona (Oct 2018)
Meeting 5 July 2018:
Tracking of OSG Tier 2's?
Track fraction of storage reachable by IPv6 per experiment (T1 and T2).
CMS Twiki notes that RAL and FNAL are not yet IPv6.
Issue of two dual-stack machines using IPv4 between them.
Meeting 5/6 June 2018:
1. FNAL and BNL Tier 1 transfers are still IPv4. For other Tier 1s - all to investigate.
2. Monitoring of xrootd transfers over IPv6.
3. KIT/dCache issues. Can Francesco & Paul Millar understand?
4. Terminate IPv6 VO.
5. More than 50% of CMS sites have IPv6 verified.
6. Perfsonar V4.1 expected in Q32018. What about all the timeouts?
7. Still need Experiment stats on weighted average IPv6 storage.
-- review of actions --
3. The KIT/dCache was discussed. Francesco had talked to Paul Millar but they had not come to a conclusion. Dave suggested writing the issue up as a contribution to the CHEP paper. Perhaps the issues BNL encountered and which caused then to revert their FTS server back from dual-stack were related? Indeed the issue caused BNL to miss the FTS deadline of last April. Also need to investigate and summarise why the other Tier1s such as FNAL and Triumf missed the same deadline.
4. Action on David to terminate the IPv6 VO.
5. Number of CMS sites (weighted by size) are now IPv6 ‘verified’. See slide 8 of Andrea’s GDB talk https://indico.cern.ch/event/651357/contributions/3128685/attachments/1714023/2765091/IPv6_deployment_update.pdf
Shawn sent details of OSG sites to Andrea for GDB last week. One LHCb Tier-2D site missed off the list of sites which were sent a ticket - problem with the site configuration in the GOCDB.
Roundtable updates
Edoardo: CERN activating a new link for CNAF. CNAF moved IPv4 but forgot to move the IPv6 link - about 50% of traffic remained on IPv6. https://netstat.cern.ch/monitoring/network-statistics/ext/?q=LHCOPN&p=LHCOPN&mn=IT-INFN-CNAF&t=Weekly
Part of moving Tier-1s (e.g. IN2P3, CNAF, KIT) to new 100G links.
Traffic: CERN last month 290PB over IPv4 and 71P over IPv6 - sum of outbound and inbound, LHCONE and LHCOPN. 71/290 = about 24% over IPv6.
Duncan: perfSOANR URL not working over IPv6 for non-LHCONE sites. Imperial and Brunel encountered a problem with EOSCMS.
Andrea: CMS glidein-WMS ITB is not quite ready. Machines at FNAL need to be made dual-stack.
Costin: Not much news. Working on moving to root6 which depends on xrootd4. Will solve client problem. At CERN 5PB of buffer that is not IPv6. So at Tier-0 80% available over IPv6, Tier-1- 54% available over IP6, Tier-2s 26%. Always use latest xrootd for data transfers.
Kars (DESY): Tier-2 readiness ticket should be fine now.
Francesco: No update from Italian sites, especially interested in Pisa. Milan moved to a new building over the summer. Also interested in the difference in speed of transfers over IPv4 and IPv6 for xrootd (from Steve Lloyd’s pages).
Bruno: still in process of getting IPv6 back online again. Changed from dual-homed to dual-stack. Reaching the end of the tunnel. Will turn on in week 41 (15th October). CTDB was an issue.
Fernando: no news. No progress on issue where two dual-stack IPv6 nodes seem to use IPv4.
Kashif: problem with IPv6 at RAL at border router - shows up as an instability. Have check internal switches etc all OK.
perfSONAR:
Marian presented status. 4.1 pscheduler - died, 4.1.1 psconfig issue. Recommend sites upgrade to CentOS7 and turn on auto-update.
CMS and LHCb ETF instances are ready, but waiting for myproxy still. Results passed to SAM3.
Tier-2 status.
See presentation at GDB https://indico.cern.ch/event/651357/contributions/3128685/attachments/1714023/2765091/IPv6_deployment_update.pdf
about 40% of T2 sites have storage accessible on IPv6
Xrootd monitoring now supports IPv6 http://xrootd.org/doc/dev44/xrd_monitoring.htm#_Toc449036991 . What will be required to deploy this in practice? Will a campaign to upgrade all the storage elements. What about the HTTP- third party copying - need to make sure that reports correctly?
Day 2 - notes taken by Francesco Prelz. IPv6 F2F meeting.
19.09.2018 Morning
The agenda for the morning is briefly reviewed.
The status of the Tier-1 is reviewed by inspecting the perfsonar mesh (notes by Dave K. here as I was tinkering with dCache - to add when I find a moment -DPK).
Issues with Fermilab: it is a large, complex site and the transition takes time.
Issues with Brookhaven: they switched off FTS after seeing problems with dcache (or even plain gridFTP) transfers. Is this only to some hypothetical rogue site on the EU site? If so, is this site reachable OK from elswehere ?
DaveK: Do we know the status of IPv6 network announces in LHCONE ? The transition GGUS tickets can be satisfied via general-purpose network as well. Do we correctly assume that all production Tier-X networks are announced in LHCONE ?
KarsO: there are roughly 200 IPv4 announces, vs. ~100 IPv6 announces on LHCONE.
Dates of next meetings:
Next F2F meeting on Thursday-Friday January 24-25 2019, lunchtime to lunchtime as usual at CERN.
Next phone meetings:
- To review CHEP paper (see below) Thursday, October 4th, 2018, at 16:00 CEST
- Thursday, October 25th, 2018 at 1600 CEST.
- Thursday, December 13th, 2018 at 1600 CET.
Planning for the CHEP paper:
CHEP has changed the publisher. It is now 'EPJ Web of Conferences'. We now retain the copyright to the paper as authors, publish under a Creative Commons license, and need to sign a piece of paper before publications.
All authors should get an ORCiD ID.
There is now a paper skeleton on the usual github project:
https://github.com/prelz/hepix_ipv6/tree/master/chep2018_paper
The author list on github is the one shown in the CHEP submission. The list is reviewed shortly.
Sections for the CHEP paper need to be identified, with volunteers assigned to write each one. Here we go:
Section 1: Introduction - including the timescale set by the management board.
Section 2: Status of the transition.
Should the experiment point of view be here or in the Introduction ?
What do the experiment reps think ?
With plots per experiment and per country.
Need to choose a cut-out date for the plots. October 1st ?
Andrea: plots are usually updated at the beginning of the month. The status won't look as good without BNL and FNAL.
This section is split in 3 subsections:
2.1 Status of the Tier-0 and Tier-1's - BrunoH is happy to edit this section. EdoardoM can provide data.
2.2 Status of the Tier-2's - AndreaS volunteers for this section.
2.3 Experiment services - CostinG gets volunteered to co-ordinate this.
Section 3: Monitoring - Should we ask Marian for this ? together with Duncan
Section 4: Future plans and conclusions - FrancescoP will draft this.
IPv6-only LHCOPN ? Some discussion on when this may start making sense.
BrunoH: Need 100% accessibility on all LHCOPN sites first.
Should there then be a section on encountered issues with dCache, etc ? Maybe part of the Tier-1 section.
A phone meeting will be held on Thursday, October 4th, at 16:00 CEST to timely discuss the status of the paper draft.
There are minutes attached to this event.
Show them.
