DIRAC Project meeting

mercredi 11 oct. 2017, 10:30 → 12:35 Europe/Paris

Visio Zoom

Johan Bregeon (IN2P3 LUPM)

The DIRAC@IN2P3 Project meeting to review the progress in various activities

• https://zoom.us/j/962565785

 

October 11th 2017 Project meeting

----------------------------
Andrei Dirac update
Dirac Release v6r19
+ same as v6r18, backward compatibility issues for old clients could not work with new server
+ few changes from the uses point of view, mostly internal changes related to computing ressources status, was used previously only for storage ressources (use RSS ?)
+ patch to come, usual business

Dirac v6r20 pre-relase triggered
+ code to handle FTS3 better now that FTS2 is deprecated
+ only new big feature of v6r20, should be released quickly
+ may be json protocole update (faster), to be discussed at the Bild meeting tomorrow

Issue with Dirac4EGI
+ many users started to use heavily the service
+ problem with sandbox side, also linked to fallback storage
  more jobs with bigger sandbox -> filled fallback storage -> crash everything...
  why did storage monitor not send any alert (mail alert developped this summer) ?
  could have the same issue for FranceGrille Dirac, need to understand the issue
  SystemAdministrator component includes an agent that checks hosts status (certificates, disk space)

----------------------------

Upgrade OpenStack + kernel security patch at CC-IN2P3
+ many different issues with many different VMs (ATLAS, CMS, CTA...)
+ problem of VM monitoring
  no graph available
  Andrei: how to couple local monitoring with DIRAC monitoring features ?
+ time issues lead to handshaking problems
  ntp service to be configured correctly, use ntp with CC router

Galera Cluster
+ change from mySam to InnoDB, we did not know
+ index +3: same problems for CTA and FranceGrille, no problem seen so far
            possible to fix, but not easy, lot of work, and may be not very efficient
+ incident because of concurrent writing on the different nodes
  everything blocked for a few hours
+ problem User Hugonie - some jobs disappeared, half of the jobs not submitted

Other issue
+ FranceGrille "complex": problem on ccdirac06 via REST API can't write sandbox on LAL machines
  to be understood
------------------------------

 

 

1 Etat du projet

Dernière version de DIRAC ?

Orateurs: Dr Andrei Tsaregorodtsev (CPPM, Marseille), Johan Bregeon (IN2P3 LUPM)

2 CTA Upgrade MariaDB et OpenStack

• Update OpenStack, and problems encounters with several servers, is that correlated ?
• DB transition from MySQL server to MariaDB Galera cluster
  ** index +3
  ** triggers and foreign keys
  ** crash

Orateur: Mme Luisa Arrabito (LUPM)

3 MetaQuery, Transformations and Productions

Orateurs: Johan Bregeon (IN2P3 LUPM), Mme Luisa Arrabito (LUPM)

4 Virtualisation and Containers

• CTA Dirac Client in a Singularity container
  ** https://github.com/ahaupt/CTA-Dirac-client
  ** singularity shell -B $X509_CERT_DIR:/opt/dirac/etc/grid-
  security/certificates ./ahaupt-CTA-Dirac-client-master.img
  
  
• VMDIRAC
• Vac/VCycle

Orateurs: Johan Bregeon (IN2P3 LUPM), Mlle Sorina POP (CNRS), Vanessa Hamar (CC - IN2P3)

5 Publication

• COMDIRAC
• Transformation System
• Meta Query

Orateurs: Dr Andrei Tsaregorodtsev (Aix Marseille Univ, CNRS/IN2P3, CPPM, Marseille, France), Johan Bregeon (IN2P3 LUPM), Pierre Gay (Université de Bordeaux)

6 Journées SUCCESS

Orateur: Luisa Arrabito (LUPM)