Page History

Date

18 May 2016

Attendees

• @AndyS @gapon @vaikunth @natepease @jgates @FabriceJammes @MichaelKelsey @AndyHanushevsky

Status Reports

@FabriceJammesFabrice: Monitoring work (

Jira
server JIRA columns key,summary,type,created,updated,due,assignee,reporter,priority,status,resolution serverId 9da94fb6-5771-303d-a785-1b6c5ab0f2d2 key DM-5885

).  Will sit with Fabien to provide monitoring reports from large scale tests.  Fixed docker git script (

Jira
server JIRA columns key,summary,type,created,updated,due,assignee,reporter,priority,status,resolution serverId 9da94fb6-5771-303d-a785-1b6c5ab0f2d2 key DM-6130

).  Command line complex, will provide docs.  Will work on Docker swarm (

Jira
server JIRA columns key,summary,type,created,updated,due,assignee,reporter,priority,status,resolution serverId 9da94fb6-5771-303d-a785-1b6c5ab0f2d2 key DM-5967

).  With Walid (student at IN2P3) working to set up scripts – can he get NCSA account?

@natepeaseNate: Butler stuff.  RFC-184 closed as adopted; working on implementation.  Storage factorization.

@vaikunthVaikunth:  Progress cleaning up Kibana dashboard, now gets only nodes we care about.  Can adopt and do analysis on data; some binning effects due to pre-aggregation?

@AndySAndy S: Finishing tech note about L1 database schema, revision, etc.  Need reviewer before publishing (soon!).  Next step to start prototyping implementation.  Biggest unknown: new L1 for each release, or accumulate forever?  Will look at MySQL vs. PostgreSQL. Must be partitioned, based on time -- newest is active; close when full and open new one.  Avoid searches across multiple partitioned.

@gaponIgor: First week; getting up to speed.  Still need HipChat account, NCSA, IN2P3 (request access to QServ cluster).  Want to build/test software before leaving.  Get VMware installed on Mac.

@MichaelKelseyMichael: Adding objectId to internal containers in lsst/partition, in order to dump secondary index.

@jgatesJohn: Threads throttled much more than expected on cluster (

Jira
server JIRA columns key,summary,type,created,updated,due,assignee,reporter,priority,status,resolution serverId 9da94fb6-5771-303d-a785-1b6c5ab0f2d2 key DM-5910

). Possible solution: use single pool of threads to allocate among requests; enforces maximum total usage.  Bigger problem: proxy gets too much data back, then dies -- eats system resources rapidly then stops, taking czar with it.  Very narrow window of "safe" queries which both exercise new code and doesn't kill proxy with results.

@AndyHanushevskyAndy H: Will discuss face-to-face w/John some changes to simplify his proxy stuff.  Logging in XRootd essentially done except for interface to QServ logger (

Jira
server JIRA columns key,summary,type,created,updated,due,assignee,reporter,priority,status,resolution serverId 9da94fb6-5771-303d-a785-1b6c5ab0f2d2 key DM-4473

).  Sequencing releases to coordinate with QServ in process.  Working on memory management stuff as well.