2024-08-23 Hyrax Fedora 6 Working Group

 Date

Aug 23, 2024@9am Eastern

 Participants

• @Arran Griffith

• Ayoub Belemih

• @Ben Pennell

  @Bradley Watson

• @Collin Brittle

• @Dan Field

• @Daniel Pierce

• @Emily Porter

• Heather Greer Klein

• @Jon Dunn

• @Juliet Hardesty

• @Kate Dohe

• @Randall Floyd

• @Rebekah Kati

• @Tom Wrobel

indicates note taker

Goals

• Updates on progress for ongoing development work

• Define and understand remaining work for Sirenia

 Discussion topics

Notes

 Action items

•  Check on if more work needs to be done to cover all the predicates from hyrax so that they get persisted to Fedora with a real URI, particularly FITS ones.

Discussed Items:

• Broken Hyrax (Pair-Tree’ed Fedora v6.5) Functions:

  • Access Control List IDs are not associating to the parent object as a value of accessTo, causing the query to find an already persisted object related to the parent to fail.

  • (From Brad) FileSets are unaware of their Work parent, since the value of proxyFor in a FileSet is not being assigned with the Work ID.

• Performance issues have been evident in Samvera’s Pair-Tree’ed Sirenia testing:

  • Creating, modifying or deleting a collection type in Sirenia takes a really long time · Issue #6888 · samvera/hyrax (This issue isn’t evident in Hyrax with a Postgres Valkyrie Adapter.)

• Question to Brad: Where is the breakdown of the value persistence occurring? A: Either Hyrax or Valkyrie, but can’t pinpoint exactly where the values mentioned above should be persisting.

• Emory University is testing a non-pair-tree’ed bare-bones Hyrax to determine if it is performant enough for their use cases.

  • They are building a collection containing 100,000 Works that each contain a FileSet with one small (~15 bytes) file attached.

  • The process of ingesting the Work and FileSet objects took two complete days, but the background jobs (attaching files to their FileSets, derivatives, etc.) will take at least until Saturday morning, so UI testing will start on 8/26/2024.

  • They have plans to deploy this code with differing test cases to an AWS EC2 for more “true-to-life” results (the code is currently running on Brad’s laptop).

• Questions about UUID:

  • Is UUID the default Fedora creation scheme when a ID slug isn’t provided? Yes.

  • The UUID is minted by Fedora, and when it writes the object to the filesystem, it then generates a sha256 of the UUID which is what is seen on disk for the OCFL directory.

  • What is the version of UUID that Fedora v6.5 is using? The out-of-the-box tool for JAVA 11, using Type 4 https://docs.oracle.com/en/java/javase/11/docs/api/java.base/java/util/UUID.html

• Anecdotal report of non-pair-tree’ed Fedora v6.5 performance: National Library of Wales reports operating Fedora storage containing 6 million objects with various ranges of child nodes attached with no performance issues reported.

• Randall Floyd: Will Brad (Emory) participate in Samvera’s upcoming Hyrax working group between 9/3/2024-9/13/2024? Yes, but only the first four days. The following week is his vacation.

• Randall will contact Brad before this working group session to discuss his findings and odd behaviors concerning NOID creation in Emory’s Hyrax v5/Fedora v6.5 application in development.

• Randall will be contributing to the Hyrax working group mentioned above.

• Who is presenting at Fedora Virtual Showcase?

  • Emily and Brad (Emory) plan to present.

  • Emory will coordinate with Daniel Pierce and/or Rebekah Kati if they present to ensure details are not repeated.