...

• Time: 2:00pm Eastern Time US (UTC-5)
• Dial-in Number: (712) 775-7035
  • Participant Code: 479307#
  • International numbers: Conference Call Information
  • Web Access: https://www.freeconferencecallhd.com/wp-content/themes/responsive/flashphone/flash-phone.php 

Attendees  

• Nick Ruest
• Andrew Woods
• David Wilcox 
• Esme Cowles

Agenda

1. Performance and scale hackathon
2. ModeShape 5 issue
3. JMeter tests with MySQL
4. Test results
5. Request on root resource with a million children fails to respond

...

Minutes

...

Hackathon

...

• University of Michigan can fund a hack house for performance and scale work
• Can hopefully schedule for this Fall, but the calendar is getting close
• Maybe in Ann Arbor?
  • Nick can probably go
  • Esme’s calendar is already pretty full

 

ModeShape 5 issue

• https://jira.duraspace.org/browse/FCREPO-2060
• Related to PostgreSQL 9.3
  • Not seeing an issue with later versions
  • No issues with MySQL
  • Probably should just recommend using PostgreSQL 9.4+
    • Esme will add this to: https://wiki.duraspace.org/display/FEDORA4x/Configuring+JDBC+Object+Store
    • Will also provide some detailed example commands for PostgreSQL and MySQL since we are pushing people in that direction
• Ingest seems to be about 15% faster with Mode5 compared with Mode4

...

JMeter tests with MySQL

• https://gist.github.com/ruebot/413ca1552ed2690654d3b1db7b347a0d
• Seems to be a null-pointer exception somewhere 
• Tested with Mode4
• Nick will test with Mode5

Test results

• Esme has done test #4 thoroughly
  • Will do test #2 with Mode5
  • Has not done test #1 due to lack of disk space
  • Currently we only have tests for #1-4
• Andrew will write test #5 and Esme will provide the file
• Esme will run test #3 until running out of disk space
  • PostgreSQL and MySQL
• Need to do some analysis of the test results
  • Are these results adequate for the kinds of use cases people actually have?
  • Can we push Fedora to its absolute limit in terms of how many resources it can contain?

...

Request on root resource with a million children fails to respond

...

• Failing to get a count of all child resources
• We can probably work around this by disabling the count
• A better solution would be to store the state, but this is could be tricky
• Or we could stop counting after x children (e.g. there are 1000+ children)
  • We do something similar in the HTML UI - we only list the first 100 links
  • You can already do this in a REST request by adding a limit
  • Andrew will implement a fix
• Is there a real use case for knowing the number of children?