Table of Contents

Time/Place

This meeting is a hybrid teleconference and IRC chat. Anyone is welcome to join...here's the info:

Attendees

Agenda

  1. Performance status and next-steps
  2. 30 seconds on Fedora 3. Maybe 1 minute.
  3. Versioning design questions and issues
    1. fcr:versions endpoint
    2. default behavior
    3. scope for impending release
  4. Admin search: expectations and current capabilities
  5. Ingesting files larger than RAM (works in some Infinispan configs but not others)

Previous Actions

  •  

Status

Osman
Esme
  • Updating benchtool to add support for performing read/update/delete operations, and allowing selecting random number generation impl. at runtime (some people report much faster performance with XORShiftRNG, I see much faster performance with java.util.Random).
  • Re-testing large-scale ingest with updated benchtool, to be able to test read/update/delete operations on large-scale repo.
  • Testing ingest of large files with different Infinispan backends to diagnose problems.
Mike
  • Spent some time determining the best way to show that versioning works in fedora 4 and how users should be exposed to it in the pre-beta release.
  • Making small fixes to enable demonstration of the feature
  • Doing some preliminary design work on the future of this functionality
Eric
  • began implementing file watcher to provide dynamic filesystem federation
  • started looking at messaging module and solr indexing schemes
  • code review and documentation
Frank
Adam
Chris
Ben

 

Scott
  • Finished documenting some test profiles  – still need to move cluster platform specs to platform profiles
  • Read up on Modeshape, Inifinspan
  • More documentation tasks (About page), acceptance testing (versioning and search)
  • AuthZ/no Authz performance testing (slow progress)
Greg
  • Working on a content model examples project, with models verified via node types API
  • Converted UNC preserved object model to a node type..
  • UNC cluster is now ready for benchmarks, test platform documented
  • Running minimal config test (single node) and setting up cluster config (4, then 8 nodes)
Ye
Andrew
  • Fedora Sponsors meeting today (x2)
    • Will be soliciting module configuration management needs/expectations
  • All fedora-dev-on-sourceforge emails have been migrated, finishing member migration
  • Move "Fedora Futures" wiki-space this weekend?
  • Received UI feedback from Robin Dean at Colorado Alliance of Research Libraries
  • Fedora website redesign, should be released next week

Minutes

Performance status and next steps

  1. Concern, having consistent approaches
  2. What are people doing
    • Esme: working on benchtool, adds read/update/delete capabilities
      • Repeating 500-GB ingest in F3 and F4
      • Looking at testing profiles in wiki
    • Scott: Has created wiki testing workflows
    • Make sure benchtool readme is complete
    • Final product: big table with profiles
      • Need to surface results and issues as soon as possible
    • Ben: ran benchtool with YourKit listening
      • Looking into addressing memory-leaks
      • Will want to share profiling results, need a download location

Fedora 3

  1. Bug report related to F3
    • There is a ticket
    • Very surprised the issue was not caught in release/integration testing

Versioning

  1. Cleaning up implementation for the next release
  2. Version endpoint currently returns RDF that is questionably useful
    • Would prefer to return an RDF list of versions
    • Need to turn what ModeShape provides into a useful response
  3. Short-term goals
    • Can see a previous version
    • Can get a previous version - FedoraContent will not currently work with versions
    • Toogle versioning
    • Can rollback - later
  4. Currently versioning is a performance hit
    • Need to note if F3 versioning is turned on in testing
    • The current F4 versioning is the "worst case scenario", maximum versioning
  5. Future: need to enable default of versioning off...

Admin Search

  1. Need to clarify current status
    • If properties are not defined in node-types, they will not be indexed
    • Default indexing dcterms
  2. Expectations
    • identifiers
    • enable indexing, but limiting access to certain search results

Ingesting files larger than RAM

  1. "file" config enables ingesting content of size larger than RAM
    • Other configs fall down
  2. Esme, Frank, and Ben working it

Status Review

  1. Greg is working on mapping F3 to F4 content models
    • Scott is also interested in this
    • Greg to create ticket

New Actions

.

#trackbackRdf ($trackbackUtils.getContentIdentifier($page) $page.title $trackbackUtils.getPingUrl($page))
  • No labels