- mgmt NeIC leaves nordforsk. PO moving to CSC. Looking for new money, for more projects, but focusing on finishing current projects. NT1 stays in Nordforsk, while investigating permanent future.(Like last year) Contracts should be circulating. Will coderefinery survive? We should be able to continue using them. They are looking for permanent funding outside NeIC. Krishnaveni has moved on to a new job. Darren taking over dcache dev. Money for ARC and dCache dev for HL-LHC should come from "somewhere else". * Trying to get rid of stuff -- BDII -- Ganglia -- SRM -- IGTF -- IPv4 Tape REST is "ready", needs to be tested by wlcg central. * New Øre Nodes Any suggestions? New switches? It would be nice to have old and new available at the same time. ACTION: Check performance of ØRE nodes and spec dcache headnodes. - SITES CSC: HIP CMS running fine. New ansible scripts. Can now be upgraded without downtimes, except pools. Thomas: It is hard to get funding for WLCG. Decision is expected next month. Then if you are on the roadmap another application must be made. UiB: Could we shuffle tape data to Bergen, so that we have more distributed free space? It is possible, but is manpower intensive. Probably too late for this year, as it will take several months and should not run during alice push to tapes. Oslo: Andrej suggest that the holes on the 21st could be caused by transfer limitations. Atlas was mostly runing monte carlo, which needs a lot of files Issues with disk performance. Caused by java only writing 8kb to disk at a time. Impossible to configure. Risk spin-locking all cpus. New versions of Java halves the problem by going to 16 kb blocks. UCPH: Maswan: Have you remembered to update the tape capacity wiki. Rune: Tapes are not labelled, but will be soon. NSC: Andrej: Has there been any discussion with Arrhenius people on WLCG usage? Jens: There is already a huge discussion on how the system should look, because the users are so diverse. Oxana: tried to push up the system, but all doors were closed. Thomas: Ask what storage system, if there is a possibility for running CVFMS there. Maswan: has tried to influence the procurement of arrhenius, but the procedure is very closed. IJS: Vega is expected to live to 2026, but might be extended to 2027. Storage will be kept until natural retirement. HPC2N: Andrej: Datastaging locking leads to long waits for files that are shared by many jobs. Another issue is remote delivery service being too slow. Issue with CAs. Cern uses non IGTF certs. Arc does not allow those CAs. Maswan suggest ARC allows multiple CA directories. UniBe: Gsciacca is happy with the new network. Everything is running as expected. - Compute ARC does not support general CAs. It would be nice to have a configurable option to join CAB and IGTF certificates. There are some policies in the way. So it must be visible to the sysadm, so that it is a conscious choice. ACTION: Send this issue for discussion in the Trust group? Options: Blend CAB and IGTF Tag unsafe downloads in job description Datastagers can failover to wget. Thomas: question to ARC 7 . Which versions of condor works with arc7. Could be problematic with old condor submissions. Maswan: Why have you not installed arc-exporter. Jens tried to install it on NSC CE on Centos 7, it did not work. Will focus on RH upgrade instead. HPC2N will try to create a virtualenv to deploy on aigert. Gsciacca: asks if UniBe should be in monitoring. Yes that would be nice. Could Vega also be enabled. Maswan: please put arc-exporter in arc7 unless it will hold up the release. -Graphs: Please make them stacked. Graphs should count cores and not jobs. ARC states and substates must be represented (e.g. inlrms:q, inrlms:r) What is the dataretension policy? 3 months in prometheus, 1 year in victoriametrics and practically infinite in a low frequency victoriametrics. - Storage see Darren's notes. - Machine Future Russia is excluded. Dubna is not. Cern agreed to not stop the extension of Dubna. No new projects with Russia. The people who were really needed from Russian were employed either at Dubna or abroad. People were afraid they would be picked up by china. So expertise issue is mostly solved. Kurchatov resources are out, mostly atlas, which is not lacking resources badly. Significant impact to alice. Tape is well below targets. So if you get new resources, please make sure it is propagated to wiki etc. Heavy ions does affect data taking. Alice takes a lot of date, atlas only basic levels. - Network - Future meetings UiB can host in the spring, beginning of May would be good. Week 12. MRichter will check. Future mmodel discussion postponed to Bergen.