Date

06 May 2015

Attendees

• Vaikunth Thukral, Serge Monkewitz, Fabrice Jammes, John Gates, Andy Salnikov, Fritz Mueller, Andy Hanushevsky, Jacek Becla

Discussion items

Large scale test

• data loaded
  • need simple statistics, like size of data loaded
• Fabrice looking into ELK (https://www.elastic.co/webinars/elk-stack-devops-environment)
• Fabrice backing up data
• problems with rebuilding stack on the build node at in2p3
  • eups uses max # of cores (it worked with 4 and 8, with 16 it is much harder to rebuild the stack)

Connections

• hardcoded limits in qserv (per worker): 32 limit for scanScheduler, 4 for GroupScheduler
• Fabrice sees no connection leak on the worker, (except on SL7)
• created story DM-2713 to check for connection leaks in integration tests

Multi-node tests prep

• Vaikunth working on loader-related issues for case 4 and 5

2605 (packaging flask dependencies)

• non blocker. Slowly, over the course of the next two months or so

1674 (data ingest)

• starting, but connection/timing problems higher priority

init.d vs systemd

• latest distros, like centOS 7.1 rely on systemd instead of init.d
• too fresh for us
• install special compatibility packages

We are mixing boost and std

• created DM-2711, DM-2712

Shared memory implementation

• ok to restrict to 2 GB