NDGF AHM --------- 24 April 2018 == GDPR and Slack === Notes not taken == SMS alarms == There's a backup in Umea in case Orestaden gets into troubles again == Perfsonar == Kind of works now (TM) at NBI and NSC, taking data for 3 weeks with a test setup (a machine for ping and another for bandwidth tests). The question is now how to keep the data. More sites should be added. Support for CentOS6 is being dropped though. The actual setup should be via OPN (close to the disks). A VM in Orestaden would be a preference. LUNARC has not got the machines, otherwise everybody in Sweden has it. Unclear whether Slovenians got suitable hardware, they should be included as well. Oslo is still waiting for the servers. == Network upgrade plans == Denmark: is in the works. Norway: 20G will be in place this year. == EGI relations == === Nagios monitoring === We have ours, and EGI has theirs, which performs worse, tests are getting stuck there and we may get tickets. This is particylarly true for ARC-CE tests. If we help them, then we can probably drop ours. The catch is, our Nagios monitors a number of our internal services, disks etc, which can not be replaced by EGI. We can probably drop the ARC-CE part from our Nagios. Some argue for the site monitoring. EGI's problem is that they have to monitor many more sites, and hit timeouts. The fetch-job test is most critical, we do it in parallel internally, and it is faster than EGI, though still gets stuck. Conclusion: probably we can't drop our Nagios, but we should be able to help EGI. === NGI-NDGF vs EGI/EOSC-Hub === EOSC-Hub requires an NGI. NGI-NDGF is a "virtual" operational NGI, not a "real" one. Still, for Norway and Denmark it might be a way of getting integrated with EOSC-Hub. == Site news == Router in Copenhagen went down meanwhile, connectivity to zanak and clom lost. === Bergen === No plans to buy new hardware, there are money but it is not yet decided how to spend them. Existing storage hardware is running out of service contract in 1.5 month, probably can get extended. CPUs probably will be provided via OpenStack (ALICE). === Copenhagen=== Got some storage funds approved for next 5 years; compute nodes also would need upgrades. Network is a never-ending story, though some progress has been achieved, 100 GB will be probably before summer, and 2x100 - later. New tape pools. Want to try ZFS, but HPC2N did not have a good experience with ZFS on Linux. === NSC === Will decommision Triolith, will be replaced by Tetralith (not for WLCG though), and many other new hardware is being bought, putting a strain on UPS. There will be a totally new cluster for WLCG. === Oslo === Abel will be operating until Q1 2020. 3 big centers are being built, one in Tromso, another in Trondheim and the third - in either Tromso or Torndheim. Meanwhile Abel will be upgraded to CentOS7 and will look into Singularity. Consider buying racks exclusively for Grid jobs, most likely AMD. Experimenting with ARM. CPUs can be provided through OpenStack, though I/O intensive bits (ARC cache) may need a special care. Plan to move ARC cache and session directories to CEPH, like Slovenians. Also got additional dCache pools on CEPH. May get ALICE tape from Bergen. === HPC2N === Run Singularity, with some limited success though. Tape migration went well. Have some UPS stories to tell, too. New hardware for WLCG will not be used for WLCG. === CSC === ALICE cluster should have been expanded but everybody forgot about it, hopefully will get extended in the end. Preliminary discussion with HIP to get more storage, and also to get some tape.