Date

13 Jan 2016

Attendees

• John Gates, Fritz Mueller, Vaikunth Thukral, Andy Salnikov, Brian Van Klaveren, Unknown User (npease), Fabrice Jammes, Serge Monkewitz, Andy Hanushevsky, Kian-Tat Lim, Jacek Becla

Discussion items

JTM

• Entire DB team plans to be there Mon-Wed.  Even Fabrice coming
• Jeff T. from Oxford coming too

docker

• running multi node integration test inside docker - query ran interactively works, the very same query ran non-interactive fails with http connection to webserv problem. Still puzzled
• finally can get core file inside dockers
• writing tech note about debugging inside docker

in2p3 cluster

• in2p3 will be upgrading the 1/2 of cluster to deal with kernel panic

mariadb

• problems with mariadb / mysql solved by switching to newest version of mysql proxy
• we are still depending on mysql client because mysql python depends on mysql, switching would impact the rest of the stack
• maybe ask SQUARE to test the stack with mariadb client?
• alternatively, when we switch to sqlalchemy that will get solved
• note that we will need mariadb for pluggable authentication everywhere anyway

shared scans

• still looking how to avoid scheduling shared scan for the same chunk on multiple replicas of that chunk
• ideas tossed around involve "relative ordering" to enforce sending all queries to "primary" copy
• Would like to send interactive queries to least loaded copy
• still thinking. Not urgent, this is just resource optimization

db/table mgmt

• document as comment in DM-2802

Planning S16 cycle

• output captured in JIRA and a spreadsheet. The spreadsheet will be published soon