Choose timezone
Your profile timezone:
Powered by Indico
v3.3.3-pre
Attended:
Alessandra (ATLAS), Latchezar (ALICE), Federico (LHCb), Pepe (CMS), Adrian , Steve, John, Boris, Mayank, Julia
Discussion after Adrian's presentation.
Latchezar asked whether IRIS is UK-centric system, Adrian confirmed.
Julia asked whether the main difference of the IRIS accounting system compared to the traditional APEL-EGI implementation is in the transfer mechanism and visualisation layer. Adrian confirmed. Julia was interested whether new implementation provides more flexibility regarding collected information. Apparently not, since the APEL client and DB part were not changed.
Discussion on improving accounting data quality
Latchezar: One week for sites to validate data is not enough. Even if we foresee longer period, there should be a possibility to change data later if required.
John: We still need some deadline for validation, otherwise sites won't do it in reasonable time
Julia: What we foresee is that after validation deadline sites won't be able to go and change data. However, we foresee a possibility to change data by the WLCG Project Office. So if site was late and there was a substantial difference between autogenerated values and site measurements, site should contact WLCG project office, which can make a change and then launch report re-generation in order to take new values into account. People agreed with this solution
Julia asked what might be a threshold for discrepancy when further investigation and follow up is required. 25% looks reasonable.
Main topic of discussion was, whether experiments agree to be involved in the investigation of inconsistencies: generated data vs local site accounting, generated data vs experiment internal accounting. Central effort has been agreed with Russian colleagues in the framework of the WLCG-Russia collaboration. Monthly reports containing problematic sites will be created. Then experiments have to be involved and follow up, in particular when comparison with the experiment measurements shows inconsistencies. Representatives of all 4 experiments agreed with the idea that experiment experts nominated for this task by every experiment would follow up on the problematic sites mentioned in the report.
Discussion after WAU presentation
Latchezar: The colour scheme should be consistent with the one used for monthly accounting reports.
Boris explained that he is planning to work on a consistent colour scheme, but it can have impact on the performance of the application
Adrian told that this feature had been implemented for IRIS accounting and he would be happy to share his implementation with Boris
Alessandra noticed lack of the UK in the country list. This is just a topology issue. In WAU UK is shown as GB as returned by CRIC. Should be fixed in CRIC
John pointed out that we should use consistent abbreviation for countries across our systems
Discussion after HTCondor accounting presentation
Mayank asked whether we need to do anything for sites which would decide to use HTCondor CE but not HTCondor batch system
Alessandra told that we should not preventively work on the solutions which might not be used.
Julia agreed that WLCG should recommend sites to go for HTCondor CE+batch combination if they decide in favour of HTCondor CE.
Julia thanked people for fruitful discussion and told that summary of this meeting would be presented at the December GDB