Difference between revisions of "IPv6 site status"
|Line 1:||Line 1:|
The status of IPv6 deployment
The status of IPv6 deployment the whole WLCG is being tracked, including GGUS tickets for each site, [https://twiki.cern.ch/twiki/bin/view/LCG/WlcgIpv6#WLCG_Tier_2_IPv6_deployment_stat here].
Latest revision as of 16:18, 30 September 2020
The status of IPv6 deployment across the whole of the WLCG is being tracked, including GGUS tickets for each site, here.
|Site||Discussed with local networking team||Asked for some IPv6 addresses||Has IPv6 addresses||IPv6 allocation||IPv6 enabled hosts (1)||IPv6 hostnames resolvable via IPv6 (2)||Joined HEPIX gridftp testbed (no longer relevant)||Joined HEPIX phedex testbed (no longer relevant)||Dual-stack perfSONAR host||Dual-stack worker nodes||Dual-stack grid services (e.g. xrootd, SRM, gridftp)||Notes||Date last reviewed or updated||IPV6 SE transfer test and state|
|RAL Tier-1||Yes||Yes||Yes||2001:630:58::/48||Yes||Yes||No||No||Yes||No||Yes||Dual stack, Squids, perfSonar, FTS, CVMFS Stratum 1 and CMS UK XRootD redirector. Echo is now fully dual stack (as of Jan 2018).||2018-09-25||Success 16-Oct-2018|
|UKI-LT2-Brunel||Yes||Yes||Yes||::/64||Yes||Yes||Not anymore||Yes (DPM)||No||Yes||Yes||All services are dual stack, including in production: Args, DPM storage, Centos7 BDII, CEs (Arc/HTC) and all worker nodes. We have an experimental pure Ipv6 node.
WLCG IPv6 ticket closed on 2017-11-06
|UKI-LT2-IC-HEP||Yes||Yes||Yes||::/64||Yes||Yes||No||No||Yes||Yes||Yes||Most services dual-stack including dCache||2017-11-07||Success 16-Oct-2018|
|UKI-LT2-QMUL||Yes||Yes||Yes||2a01:56c0:4033::/48 (for grid cluster)||Yes||Yes||No||Yes (Storm)||Yes||two nodes in test with NAT64||Yes||dual-stack VLAN which also does jumbo frames. RIPE atlas probe, perfsonar and all grid services (CEs, SEs, xrootd, bdii, argus, squid) available via dual stack.||2018-12-11||Success 16-Oct-2018|
|UKI-LT2-RHUL||Yes||Yes||Yes||2001:630:113:3::/48||Yes||No||No||Working dual stack perfsonar exists. IT stuck with IPv6 reverse DNS lookup, which is needed for perfsonar mesh participation.||2018-09-18||No|
|UKI-LT2-UCL-HEP||Yes||Yes||Yes||2001:630:106:1100::/64||Yes||No||Yes||Dual-stack PerfSonar. No storage at site.||2018-11-06||N/A as no storage at site|
|UKI-NORTHGRID-LANCS-HEP||Yes||Yes||Yes||2001:630:80:2fd::/64||Yes||Yes||No||No||Yes||No||Yes||All storage dualstacked, onto squids next. Compute won't be dualstacked in the foreseeable future.||2018-07-09||Success 16-Oct-2018|
|UKI-NORTHGRID-LIV-HEP||Yes||Yes||Yes||2001:630:21:8d80::/64||Yes||No||No||No||Yes||No||No||Addresses used on perfsonar systems; waiting for sign-off for final IPv6 plan from IT services.||2018-07-11||No|
|UKI-NORTHGRID-MAN-HEP||Yes||Yes||Yes||2001:630:22:1004::/64||Yes||Yes (but no reverse lookup yet)||No||No||Yes||No||Yes||The IPv6 subnet may or may not change in the future. Services on IPv6: testbed CREAM CE, IPv6 only GridPP VOMS daemon (voms6.gridpp.ac.uk)||2018-09-18||No|
|UKI-NORTHGRID-SHEF-HEP||Yes||Yes||Yes||No||No||No||No||No||No||No||The University of Sheffield now has ipv6 enabled. The plan is to start with perfsonar. It will be dual-stack by the middle of November.||2018-10-30||No|
|UKI-SCOTGRID-DURHAM||Yes||Yes||Yes||2001:630:a5:1200::0/64||Yes||Yes||Yes||Yes||Yes||Dual Stack Personar, Nodes and Storage. IPv4 only CE/BDII/ARGUS.||2019-08-16||Yes|
Initially we migrated just our storage head nodes to IPv6. This exposed problems with DPM. This was motivated by concerns by our site network team about the path that IPv6 packets left the university and the possibility that they wouldn't be directed through Janet. After being ticketed and discovering the problems with DPM in a mixed state we have reverted our roll out in production back to just IPv4. We have our test storage node and disk node dual stacked for testing and are currently engaged in conversations with our networking team as to when is best to dual-stack our production storage. This is ongoing due to staff holidays as well as a pending major networking intervention later in the year on the site. We have started speaking with the administrators who run the RDF facility about IPv6 and will update once it's clear what the restraints may be over doing this.
|UKI-SCOTGRID-GLASGOW||Yes||Yes||Yes||2001:630:40:ef::0/64||Yes||Yes||Yes (left to join Phedex)||Yes (DPM)||Yes||No||No||Initial introduction to IPv6 for production system, PerfSonar now dual stack, DNS and Reverse DNS confirmed to work and accessible via IPv6. Issues with inbound traffic have stalled progress awaiting Central IT to resolve.||2018-10-9||No|
|UKI-SOUTHGRID-BHAM-HEP||Yes||Yes||Yes||2001:630:1dc:1100::/56||Yes||No||No||No||No||No||No||New IT DNS system in place. Can now request IPv6 DNS names. Plan to get this tested early 2019.||2019-01-22|
|UKI-SOUTHGRID-BRIS||Yes||Yes||Yes||2001:630:e4:2810::/64 2001:630:e4:2811::/64||Yes||Yes||No||No||No||Yes-SL7-WN-only||Yes||IPv6 enabled on WNs. XrootD access fails if WN's IPv6 is not added to DNS or /etc/hosts.||11 Dec 2018||Success 16-Oct-2018 (xrootd only!)|
|UKI-SOUTHGRID-CAM-HEP||Yes||Yes||Yes||2a05:b400:104:100::/56||Yes||Yes||No||No||Yes||No||Yes||New address range in RIPE /32 allocated. IPv6 now enabled on network. PerfSonar dual-stack from 2018-09-21. Storage dual-stack on 2018-11-08||2018-11-08||Success 20-Nov-2018|
|UKI-SOUTHGRID-OX-HEP||Yes||Yes||Yes||2001:630:441:900::/56||Yes||No||No||Yes||No||The IPv6 service from the University now runs on new hardware that supports it natively. However they do not provide DNS servies so we're only deploying test systems, no production work is being run over IPv6. Oxford Central IT Services team is looking into it||2018-08-14||No|
|Yes||Yes||No||No||Yes||No||No|| New IPv6 Network Configuration working PerfSONAR nodes is dual stack plus 2 test boxes
We deployed IPv6 across dCache on 2018-11-15 but as the IPv6 load through the router ramped up it started dropping packets and had a very long RTT through the router for IPv6 . We rebooted the router on 2018-11-20 which appeared top solve the problem and it looked like everything was working but then overnight it developed the same problem. We have backed out the deployment on the storage and are planning to upgrade the firmware on the router to the latest version next week after which we will try again
|UKI-SOUTHGRID-SUSX||Yes||Yes||Yes||Overall UoS range:
are in a private network)
|Yes||Since the end of August our service nodes, Cream CE, StoRM and Perfsonar, are dual stacked, the ticket was closed accordingly.||2018-09-15||Success 20-Nov-2018|
(1) Hosts with AAAA records exist in the public DNS, e.g.
dig AAAA netmon00.grid.hep.ph.ic.ac.uk
netmon00.grid.hep.ph.ic.ac.uk. 300 IN AAAA 2001:630:12:580:207:43ff:fe11:ffb0
(2) Hostnames can be resolved from an external IPv6 only host, i.e. local DNS server has AAAA record, e.g.
ns0.ic.ac.uk. 86400 IN A 188.8.131.52
ns0.ic.ac.uk. 300 IN AAAA 2001:630:12:600:1::80