Choose timezone
Your profile timezone:
HEPiX IPv6 working group meeting
→
Europe/Zurich
Vidyo
Vidyo
Description
The timings are approximate!
Vidyo meeting of the HEPiX IPv6 Working Group
7th March 2019
Notes by Dave Kelsey
Present: CatalinC (RAL), FrancesoP (Milano), AndreaS (CERN/CMS), DuncanR (Imperial), RajaN (RAL/LHCb), BrunoH (KIT), AndreaM (CERN/FTS), DaveK (RAL/Chair)
Apologies: EdoardoM (CERN), FernandoL (pic), ThomasF (DESY), MarianB (CERN), MartinB (RAL).
1. Agenda was agreed. No actions to review.
2. Tier 2 Status.
AndreaS takes us through his standard tracking plots.
https://indico.cern.ch/event/797495/contributions/3313608/attachments/1808147/2951906/go
Good news is that for the first time two large countries have completed their Tier-2 transition, namely France and Spain - congratulations!
Overall nearly 60% of sites and just over 60% of the storage capacity is IPv6-capable.
3. Tier 1 Status
BrunoH presents recent news. DuncanR informs us of hot news (less than one hour before) - BNL Tier-1 has re-enabled IPv6 on their FTS server and transfers have started.
FNAL Tier-1 is still working on their deployment.
There are local problems at TRIUMF, it seems, blocking IPv6 transfers in from KIT - a firewall issue?
4. HEPiX abstract submission.
DaveK presented his draft abstract for the HEPiX meeting in San Diego at the end of March 2019. A number of improvements were discussed and agreed. DaveK will distribute a new version later and submit on 8th March.
5. Roundtable updates.
RajaN reports that LHCb jobs are now running again on Brunel IPv6-only WNs. Problems needed fixing on both the CE and the system receiving logs.
FrancescoP reports that he has produced a patch for dCache relating to manual configuration to fix the known problem with multiple network interfaces. He is looking for a site willing to test this - BrunoH will check at KIT.
DuncanR reports that he helped debug problems with ATLAS jobs on IPv6-only WNs at Brunel - the fix required changes to the Rucio data transfer client - changes made but not yet tested. He also reports problems transferring data to TRIUMF. The GridFTP port needed to be opened (is this same problem as noted above between KIT and TRIUMF?).
6. FTS, monitoring and dashboards.
AndreaM reported on FTS actions since the January F2F meeting. Problems had been found involving Florida and FNAL (from CERN) requiring IPv6 to be disabled on FTS. This was fixed and re-enabled a few weeks later. He also reports the very recent FTS IPv6 enabling at BNL, but notes that there is not yet much traffic, partly because the number of dual-stack US ATLAS Tier-2s is small. He will continue to investigate and check that dual-stack is configured correctly and that IPv6 transfers are happening.
MarianB in sending his apologies for the meeting had reported two things:
a. perfSONAR situation has improved a lot, we’re now measuring IPv6 as planned, thanks to Duncan we have also made quite good progress with maddash (http://psmad.opensciencegrid.org/maddash-webui/index.cgi) and pushed most of the T1s to fix their perfSONAR nodes (including RAL) and we have also managed to found few bugs that were fixed by the developers. Remaining T1 to be fixed are KISTI, TRIUMF and CERN (here we’re waiting for security to scan the new nodes and open central firewall). The focus will now turn towards major T2s as there we have also a lot of configuration issues to fix.
b. As for ETF, ATLAS has asked for ETF IPv6 instance and there is now one running at https://etf-atlas-ipv6.cern.ch/etf/check_mk/ - and it’s IPv6-only! We have made good progress in improving the tests as most of them were failing once we moved from dual stack to IPv6-only and while there are still some issues to fix, ATLAS seems to be very happy with the progress. ATLAS also plans to introduce DNS AAAA test as was done by CMS some time ago. I plan to migrate LHCb and CMS instances to IPv6-only soon.
DuncanR also talked about the current situation with perfSONAR and showed several maddash results. He had been chasing Tier-1s anyway about upgrading to CENTOS 7 and this was a good opportunity to solve other problems. NDGF has problems with the Latency measurement and we discussed the situation re FNAL, where BrunoH reported on the various PS systems in use their, only some of which are IPv6-capable.
DuncanR also reported that he continues to chase the xRootD monitoring of IPv6 traffic - this is getting nearer to being fixed. Some more testing (and volunteers) are required.
7. AOB
DuncanR reminds us that we do need to find an ATLAS representative for this working group (to replace AlastairD).
Dates of next meetings
Thursday 11th April - Vidyo starting at 16:00 CEST.
Next F2F meeting at CERN - Thursday 2nd/Friday 3rd May 2019.
(see agenda entries on the Indico group https://indico.cern.ch/category/3538/
Meeting closed at 17:00 CET.
There are minutes attached to this event.
Show them.