2022-07-27 Meeting notes

 Date

Jul 27, 2022

 Participants

  • @James Walder

  • @Ian Johnson

  • @Alison Packer

  • @Thomas, Jyothish (STFC,RAL,SC)

  • @Emmanuel Bejide

  • Lancaster: Gerard

  • Glasgow: Sam

 Goals

  • List of Epics

  • New tickets

  • Consider new functionality / items

  • Detailed discussion of important topics

  • Site report activity

 Discussion topics

https://stfc.atlassian.net/jira/software/c/projects/XRD/boards/26/roadmap

Item

Presenter

Notes

Item

Presenter

Notes

5.4.3 releases in Centos 7 / EL8;

  • Problems observed: (pgRead / pgWrites)

@Thomas, Jyothish (STFC,RAL,SC)

https://stfc.atlassian.net/browse/XRD-14 , https://stfc.atlassian.net/browse/XRD-10

Identified the issue, client readv calls exceeded 1024. @Thomas, Jyothish (STFC,RAL,SC)
prepared a server fix. Michal (XrootD dev) prepared a client side fix.

Outstanding items for 5.4.X XrdCeph version

 

Master: https://stfc.atlassian.net/browse/XRD-22 included;

 

BufferedIO: PR to merge current Master needed;
https://stfc.atlassian.net/browse/XRD-9
Code prepared; PR needed

https://stfc.atlassian.net/browse/XRD-24

 

Add xrd.report monitoring to all xrootd instances; collate and ingest into InfluxDB.
Initial version: https://github.com/snafus/xrdreport

 

 

https://stfc.atlassian.net/browse/XRD-26

 

ceph-dev-gw2: passing current compliance tests (XrootD 5.5. will be needed for all components).
Demonstrate that Echo can satisfy proposed ATLAS strategy with mixed x509 / tokens workload.

 

Deletions updates

@Ian Johnson

Expanded deletion time measurement script to automate more of the recording. Running measurements on nGB-size files (up to 8GB). Deletions of such files take < 2s for me. (Problems uploading the test data though - for 8GB file, upload takes ~ 40 minutes or errors such as “(Neon): Could not send request body: Connection reset by peer” after 20-30 minutes…)

 

 

 

 

Brief Summary from discussion with LHCb / Chris

  • gfal2 doesn’t support timeouts for removal ( to check and send any instructions to Chris? )
    Write conflicts with different protocols: Dirac will immediately try another protocol.
    To switch to davs once new gateways are up.

  • Rob C’s code is the measure of performance for VR. User jobs have various counter-measures in place.

gridFTP for transfers for RAL and other TPC transfers in the UK. Some uk sites only supporting gridFTP? (to understand more details).

 

Site reports

Oxford Xcache:
- Xrootd service failed to start due to failed DNS name resolution of the monitoring host specified in xrootd.monitoring directive.

 

Lancaster:

xrdcp TPC working back again.

 Action items

Reminder from Sam to update GGUS tickets

 Decisions