Attendees
General
- Indicates who took minutes -
- Call-in: Google-hangout at:
Goals
- Top-level: Get to 4.0-Beta by Apr 1
- Sprint-level:
- Clustering
- Ramp-up new team members
- Wiring and Config
- Durable persistence
- Unit test coverage > 75%
- Bug fixes
- Prepare for 4.0 feature sign-off
Team Focus
- Greg Jansen
- Issues related to AuthZ in Modeshape 3.7.x upgrade
- Clustering with Frank
- AuthZ unit tests
- Ben Pennell & Mike Daines
- Read docs: https://wiki.duraspace.org/display/FF/Project+References
- Get dev environment up
- AuthZ unit tests with Greg
- Kernel API refactoring with Adam
- Scott Prater
- Feature sign-off/documentation
- Clustering with Frank
- Osman Din
- Persistence inspection utility
- Eric James
- Indexing with Adam
- Adam Soroka
- OSGi
- Kernel API refactoring (Jena)
- Kernel refactoring (update pattern)
- Mike Durbin
- Transactions
- Feature sign-off
- Much, much more
- Frank Asseg
- Clustering/performance
Minutes
Logistics
Note: Hangout is capped at 10 and we have 10 this sprint. ReadyTalk (or similar) will be used for committer calls
Frank out sick today
Goal this week: have decent headsets with quality mics for stand ups
Introduced Mike Daines and Ben Pennell from UNC
Project process review
wiki: Andrew will make an account for Mike
pivotal tickets:
invites sent to Mike and Ben
wiki page on process was edited
ideal day ~= 3 pts on pivotal
scope tickets to a day or two
link to PR, click finish, start review/acceptance
getting PRs in early helps reviewers
Goals
Overall Goal: Get to 4.0-Beta by Apr 1 (Q1), 4.0 release at end of Q2
Beta is defined by feature completeness, agreed across the community
Most features are now there in some form, significantly complete
Clustering is one area that is less complete and needs work
Sprint Goals
Clustering
Ramp-up new team members
Wiring and Config
various options in codebase, e.g. authz, projection/federation
not easily chosen or configured at this time
Durable persistence
communicating the position of the project to the community and getting feedback
elaborate options
Unit test coverage > 75%
perhaps also IT coverage, though we do better there.
IntelliJ will do coverage out of the box, identify hotspots
Jacoco does not require network access, Sonar phones home
Clover is another option (busy interface)
we have an IntelliJ license via the project, ask for one if needed
Bug fixes
several fixes, some coming from user community
Prepare for 4.0 feature sign-off
internally evaluate feature completeness (and docs) and get ready for external stakeholders to accept/sign off on features
An Aside on New Tickets:
Several tickets have already been assigned, but new tickets still need to be created..
Through the course of the sprint and this meeting, making a ticket is encourage as the need arises
Walk through by goals
Clustering:
UNC - cluster ready to go
Scott - Wisconsin cluster ready to go (getting keys today)
Frank
As admin you can set up a cluster of X nodes, probably a load balancer
work through recommended Infinispan configs
Able to perform normal operations in a performant, scalable way
surface errors in basic functionality
verify ingest/delete operations are consistent
UNC and Wisconsin need to help define what is performant
improve upon F4 single node, and therefore F3 single node
specific institutional requirements
Frank has some specific rates of ingest
Adding nodes/resources should increase performance
Need tickets, not many tickets in pivotal yet
X number of owner nodes for any object (distributed to x level)
use benchtool to measure performance
compare performance across cluster sizes
https://github.com/futures/fcrepo-aws-puppet for installing all requires on a basic linux install
there is a load balancer config here: https://wiki.duraspace.org/display/FF/ModeShape+Clustering
Ramp up Mike and Ben
some tickets might relate to underlying technologies, reading up
getting environment up and running
build code
do quick start
deploy WAR file to container
IDE setup
extracting the public Kernel API from kernel module, Andrew to provide guidance
Unit test coverage
filling these gaps is a good way to become familiar w/project
authz needs more unit test coverage
sonar may point at other gaps to reach 75% goal
current coverage is 64.5% unit tests, 71.4% integration tests
Mike Daines and Ben, create tickets for writing unit tests (perhaps by class)
Wiring and Configuration
JMS indexer, progress on driving problem, Adam would like to finish indexer modularization for OSGi framework
drop in a new indexer at runtime (hot deploy directory and web admin UI)
Other A. Soroka topics:
- New mechanisms for property updates
issues with Jena machinery behind SPARQL
Frank reported no issue with heap and SPARQL, therefore this can be lower priority
- More conversation and definition needed for mapping from JCR to RDF notions
- cohesive description of resources
Durable persistence
Osman’s utility provider visibility into files that land on disk
under most configs Binary JSON show up on filesystem with all info, but not really intelligible
utility reads in a file/dir, then exports the contents in a user friendly format
goal is to get this packaged and available such that users can inspect their persistent store.
Andrew will communicate this persistence model to the community, gather feedback
Transactions
Mike Durbin needs the patch for this, which was not attached to email (Andrew will forward)
transactions are broken in many common scenarios
An Aside regarding Mike Durbin
- Various bugs have been identified and assigned in pivotal
Missing Graph info about children - assigned Mike Durbin
Preparing for Sign off on 4.0 feature set
Andrew and M. Durbin to look at feature set as comparing to wiki roadmap
find gaps and do thorough testing
perhaps spin off additional documentation tasks for other team members
target stakeholders for sign-off
Versioning
reverting to a previous version is missing, there are tickets existing, M. Durbin to continue working these.
Going through Pivotal Tracker
Upgrade to Modeshape 3.7.1 (GJ)
includes authz work/fixes
Transactions (M. Durbin or someone else)
patch might do it
Finding new limits of describe functionality, testing interator/streaming child implementation for heap fix (Ben or Mike Daines, or Durbin)
Get in touch with Mike Durbin for new work (he has extra)
Please create new tickets by stand up on Tuesday, the we can prioritize in the meeting