GridPP Technical Meeting - PerfSonar 5

Friday 9 Jun 2023, 11:00 → 12:50 Europe/London

Virtual Only

Alastair Dewhurst (Science and Technology Facilities Council STFC (GB)), Andrew McNab (University of Manchester), David Colling (Imperial College (GB))

Weekly meeting slot for technical topics. We will try and focus on one topic per meeting.  We will announce at the Tuesday Ops meeting if this meeting is going ahead and if so the topic to be discussed.

 

 

Check MK shows the status of UK sites:

https://psetf.opensciencegrid.org/etf/check_mk/index.py?start_url=%2Fetf%2Fcheck_mk%2Fview.py%3Fhostgroup%3DUK%26opthost_group%3DUK%26view_name%3Dhostgroup

 

 

New monitoring based on Grafana is being currently developed.  Duncan thinks it will be avalable in Late Summer.  Duncan provided some links:

https://monit-grafana-open.cern.ch/d/-SMRsAx7z/perfsonar-h2h-performance?orgId=16&var-src_bw=ps-london-bw.ja.net&var-src_lat=ps-london-lat.ja.net&var-dst_bw=lhcmon3.bnl.gov&var-dst_lat=lhcperfmon.bnl.gov&var-dst_bw1=perfsonar-test-v1683-portal.gridka.de&var-dst_lat1=perfsonar-latency-v471-portal.gridka.de&var-dst_bw2=lt2ps00-bw.grid.hep.ph.ic.ac.uk&var-dst_lat2=lt2ps00-lat.grid.hep.ph.ic.ac.uk&from=now-24h&to=now

 

Feedback from sites:

- Lancaster

perfSonar is ok, but it is a black box.

Shame that perfSonar 5.0 doesn’t support more modern OS.  – Tim Chown added that EL8 support will be available any day now and EL9 within a few months.

 

 

- Glasgow

Upgraded to perfSonar 5 automatically and appears to be working.  It gets very little attention unless someone pokes them.  Did hit the logs not rotating problem.

 

 

- Oxford

Upgrade happened automatically.  Something went wrong and Duncan had to contact them.  It didn’t work for 3 days and then started to work.  Vip feels there are too many logs produced that makes debugging actually harder.  Vip also wonders why service restarts don’t work and he has to reboot machines.  Raul doesn’t believe this is necessary but notes that there are several services perfSonar depends on that might been to be restarted,

Support other site comments about

 

 

- Brunel

PerfSonar host is too old and in the process of purchasing a new machine.

 

 

- Imperial

Experimenting with minimal testpoint install as they are concerned that the automatic upgrades and enabled features that potentially opened things up to the internet.

He way that the RPM works is “brittle” and the post install scripts are not easy to work with.

 

 

- ECDF

Update to perfSonar 5 went smoothly.  Hit same log rotate bug as others.  Also found that pschedule claims it wasn’t running when it was.

11:00 → 11:20 PerfSonar 5

Speakers: Duncan Rand (Imperial College (GB)), Raul Cardoso Lopes (Brunel University (GB)), Tim Chown

11:20 → 11:45 Site Round Table

Speaker: Alastair Dewhurst (Science and Technology Facilities Council STFC (GB))

11:45 → 12:00 AoB