Choose timezone
Your profile timezone:
HEPiX IPv6 working group meeting
→
Europe/Zurich
Vidyo
Vidyo
HEPiX IPv6 working group meeting - Vidyo - Thursday 9 July 2015
Present: Alastair Dewhurst, Thomas Finnern, Terry Froy, Costin Grigoras, Kashif Hafeez, Thomas Hartmann, Bruno Hoeft, Tiju Idiculla, Dave Kelsey, Raja Nandakumar, Kars Ohrenberg, Francesco Prelz, Andrea Sciaba, Ulf Tigerstedt, Dan Traynor, Ramiro Voicu.
Apologies: Fernando Lopez Munoz.
Actions and ongoing issues
- DaveK. Ask for news from ALICE/CostinG. Done. Costin is here today.
- AndreaS. Dual-stack SAM3 monitoring. AlastairD previously reported that in ATLAS something is dual-stack on SAM. AndreaS will follow up. Ongoing.
- DaveK. Security Best Practices document. Ongoing.
- FrancescoP. Setup for XRootD testing. Previously started by recompiling XRootD version 4.2.1. Worked on reference FAX installation. Then had to remove ATLAS related components. He needs a few servers. Ongoing.
Roundtable updates
a) PIC - Fernando sent apologies for today but reported that they have now connected to LHCONE via IPv6.
b) INFN. Francesco reported that GARR now gives a default route. He needs to do a test with other remote servers. He tried to PIC and CERN but this failed. It was noted that the dCache testbed nodes are also gridFTP end points. He tried some simple client - these took 5 to 6 seconds using IPv4 and IPv6 addresses. Just one day later it now only uses on IPv6 address. Seems to have been a problem for whole of first session. Also FP now has a dual-stack perfSONAR node. He needs to inform Bruno and Duncan.
c) ATLAS. Alistair says there is little to report. QMUL has been setting up dual stack CEs where there are problems with IPv4 pilot jobs working on the dual-stack CEs. There are inadvertant GGUS tickets related to test SE endpoints (Glasgow and Oxford as before).
d) KIT. Bruno is following up on LHCOPN/LHCONE and IPv6. ThomasH reported on the tests he did with FTS etc. In the end it was not gfal2. globus was the culprit. Fixed now and 29 transfers have succeeded in the last hour.
After the meeting, ThomasH sent this update by email:
I got gridftp transfers to work again with all packages updated to their recent versions but with globus-ftp-client downgraded from 8.21 to 8.19. The corresponding ticket is https://globus.atlassian.net/browse/GT-604
On a quick search I did not found changelogs for globus-ftp-client packages (only for major releases in http://toolkit.globus.org/toolkit/advisories.html )
Digging into the gridftp affecting tickets over the last months ticket https://globus.atlassian.net/browse/GT-568 looks quite interesting. It is about failing transfers between a IPv4-only and a dual-stack end point... so the other way round.
e) ALICE. Costin reports that their clients have been updated to XRootD v 4.1.x and these are now used for all data management. Third party transfers have problem with IPv6 where they try to use the same protocol on the whole chain.
f) QMUL. Dan reports that perfSONAR boxes are now dual-stack as are test and production StoRM SEs. All with no problems. The CREAM CEs are moving to new hardware on dual-stack. The last one was moved at the same time as an aircon failure since when no jobs are running. ATLAS pilot job factories (IPv4) are now failing.
g) DESY. Kars says nothing much to report. Yesterday Deutsche Telekom switched on IPv6.
h) RAL. Tiju reports he is still working on a S3 gateway. This works on IPv4 but not IPv6. They are still battling router problems with the vendor (nothing to do with IPv6).
i) LHCb. Raja reports that the IPv6 library problem reported earlier has now been fixed at CERN and now tests work (for last couple of weeks). The IPv6 version of DIRAC is now used by other VOs and IPv4 has not been broken. QMUL has submitted a ticket for 2 dual-stack CEs where LHCb work is no working.
j) Nordic. Ulf reports that they finally have all dCache nodes on IPv6. There is still one backup link in Norway not routing IPv6 correctly.
Testing plans
TonyW was not present so we are not aware of the current testbed status.
Re XRootD testing. We are looking for a volunteer with a client node to contribute (does not need to be a powerful machine). QMUL (Dan) reports he has a test system. To be followed up at the F2F meeting.
Dates of future meetings
F2F meeting at CERN Tuesday/Wednesday 15/16 September 2015 at CERN (lunchtime to lunchtime).
Notes by Dave Kelsey
8 September 2015
There are minutes attached to this event.
Show them.