Integration of the new benchmark in the accounting workflow

Friday 18 Oct 2024, 11:30 → 12:00 Europe/Zurich

Zoom link:

https://cern.zoom.us/j/63307717078?pwd=OXFSeWJPUi90V0IxUHVrOFBhZVpMZz09

 

 

Attended:

Alessandro, Adrian, Carlos, Julia

Excused :

Joseph

 

APEL

Adrian confirmed that APEL does read GLUE2 and that changes in APEL client would be required.

Still need to understand how sites which publish only GLUE2 are being processed by APEL.

Need to check case by case. Julia will prepare a list for Adrian to investigate.

Unless APEL is capable of reading GLUE2, there is no point to push sites to publish GLUE2. However, we need to fix sites which publish with both GLUE1 and GLUE2 and do it not consistently.

We need to enable a possibility to keep in APEL and retrieve from it a complete list of behchmarking factors (independently of how benchmarking factors were published including local APEL configuration) used by APEL in order to compare it with values obtained by benchmarking framework from the tests.

OSG is investigating cases when OSG sites do not properly publish CPU and as a result their CPU efficiency is 0.

Next meeting is Friday the 22nd 11:30 CET

 

 

 

11:30 → 11:50 Introduction

Speaker: Adrian Coveney (STFC UKRI)

Attended:

Adrian, Alessandro, Renato, Carlos, Ivan , Joseph, Julia

 

Adrian reported about progress of APEL service migration, current step is DB migration. After completion main focus will be given to deployement to production of the APEL server with support of HEPSCORE functionality. The time line to have it in production y the end of November should be respected.

Julia asked Adrian how APEL translates time to work for sites which publish benchmarking factors using GLUE2 , while according to the code, GLUE2 format is not considered. The investigation of people working in the benchmarking project demonstrated taht among at least ATLAS sites there is plenty of sites which are publishing only with GLUE2 (see slide 17 of the presentation https://indico.cern.ch/event/1462197/contributions/6159281/attachments/2940215/5165207/wlcg_operations-analysis_corepower-20241003.pdf). Adrian promised to check APEL code. 

Julia told that according to the investigation mentioned above we are in a quite messy situation. Some sites are publishing with GLUE1, some other are with GLUE2, some with both but the numbers do not always agree. We (in this context EGI and WLCG ops) need to consolidate info in BDII, in order to do it properly we need to understand what APEL is doing regarding benchmarking factors. For sites which publish benchmarking factors in local configuration, we need a way to collect and publish them centrally in a single place. For WLCG site WLCG CRIC can be an option. But we need a way to get thses numbers form APEL in some way.

Julia asked Adrian to address two different issues reported from the sites - OU_OSCER_ATLAS and  WEIZMANN-LCG2, people are not getting answer for a while

Carlos and Ivan reported the EGI - QUANTA contarct was finally signed. Ivan tols that QUANTA started to actively work on the portal, person power is allocated. Weekly porgess reports will be sent to the lilst before our weekly meeting. There will be a dedicated mailing list of QUANTA. Issues and requests will be tracked via jira.

Next meeting is next Friday the 25tht of October same time slot.