You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@drill.apache.org by Jason Altekruse <al...@gmail.com> on 2016/02/24 22:52:00 UTC
Hangout notes from the last few meetings
Hey guys,
Sorry I haven't been sending these out, I keep meaning to go back to them
and clean them up before sending them out and I don't get around to it. I
will just post the raw notes after the meeting going forward and provide
clarification on the thread if anyone has questions.
Drill Hangout - 2/9/2016
- Attendees: Yulia, Sean, Vicky, Sudheesh, Neeraja, Karol Potocki, Arina,
Aman, Hakim, Jinfeng
- New community members, Welcome!
- Arina
- working with MapR team
- Karol
- tiny contribution to allow spacial queries in Drill
- interested in sparking interest in geo locations
- PR outstanding for shapefile format
- Neeraja - would be nice for simple doc for users to start
- examples in PR and Karol's github repo
- he could write a blog post for the apache repo
- Discussion topics
- Sudheesh
- 4281 - client impersonation
- post a design doc soon
- some drill deployments
- tableau desktop is presentation layer on top of Drill
- users only use tableau desktop, talking through tableau server
- want to pass user from tableau desktop through the tableau
server
so that impersonation works correctly
- requires a change in Tableau as well, working with the team
there
- Yulia
- 4132 - simple queries in parallel
- design doc on JIRA
- 2 goals
- separate planning from execution
- separate fragment plans so that they can be run independently
- those available please review design doc and PR
- 1.5.0
- new vote out soon
- Jinfeng
- 2517 - directory pruning in calcite logical
- vicky seems to have found a bug
- follow up work
- need to separate the rules and run them individually to
improve planning performance
- Drill user survey
- other projects list who is using them
- just a google survey
- simple questions, I assume all will be considered optional
- current drill version in use
- cluster size
- datasources used
- clients: sqlline, REST, Applications, JDBC, ODBC, BI Tools
- what is your use case?
- why Drill?
- data formats, data types
- are you using any of the security features of Drill to
restrict access of some data to users?
- view chaining, impersonation, Web UI security
- SQL features you would like to see as enhancements soon?
- how many users are querying your Drill cluster
- have you written a storage plugin, UDF or format plugin?
- issues with the build
- jdbc-all jar size enforcement
- jacques made changes to remove proguard and generally fix up
jdbc-all JAR
- 1.4.0 has a large JDBC-all jar that wasn't excluding what it was
supposed to
- Aman
- Dechang - perf regressions on rc2 metadata cache
Drill Hangout - 2/16/2016
- Attendees: Parth, Andries, Arina, Jason, Vitalii
- Topics for discussion
- Release
- issues with publishing the web site
- annoucnement should be up shortly
- Jacques had mentioned Metadata caching
- follow up if he wants to post thoughts
- Discussion was short today
Drill Hangout - 2/23/2016
- Attendees: Jason, Minji, Laurent, Arina, Parth, Sudheesh, Zelaine
arina -- modify calcite, timestamp related function --> contact calcite
folks/julien
improve c++ client, better distribution of queries across cluster,
randomization routine not distributing uniformly.
session options not allowed since can't maintain sessions if uniformly
distributed
--> c++ client std c library rand() function not always good
--> different random number generator
--> new connnection in the pool, then need to keep track of all the
altersessions (temporary tables, new schema, etc.)
--> small number of clients, need foreman workload distributed more
(planning and so on)
--> ping jacques
impersonation--> client to impersonate other clients (Delegation?)
--> odbc/jdbc: provide an api (c++/java) and how they will use it
--> waiting on comments
better testing for operator: better tests for independent components
--> mock internal parts of systems
--> run operators in isolation (posting soon)
--> exchanges needs a bit more discussion (vector container) - separate way
to mock data coming in
juliens test changes to run tests on multiple drill bits (?)
--> This actually wasn't Julien's contribution as was in the meeting,
Sudheesh was actually referring to Andrew's PR here:
https://github.com/apache/drill/pull/135