2018-04-05 - System Operations and Management SIG Notes

2018-04-05 - System Operations and Management SIG Notes

Date

Apr 5, 2018

Attendees

  • @Chris Manly

  • @Michael Winkler

  • @patty.wanninger

  • @Brandon Tharp

  • @Tod Olson

  • @Mike Gorrell

  • @spampell

  • @Craig Boman

Goals

Discussion items

Time

Item

Who

Notes

Time

Item

Who

Notes

Administration

Chis

  • Note taker for this meeting - @patty.wanninger

  • Agenda items

 5 min

Review action items

Chris

 



WOLFcon topic review



  • Monday session = hands on rather than demo, hackathon. @Christopher Creswellmentioned that they will use VMWare to create multi-node sessions during deployment. Cornell will use AWS. TAMU VMWare. UChicago is keeping all options open. The reference platform will be created with/for VMware, and AWS users will spin off a "Here's how" document.

  • Tues - Data migration process flow, conflicts with early implementers am meeting(they also meet in pm)



Discuss install experiences



  • TAMU is looking at orchestration solutions to install without OKAPI. Everything from Docker Swarm to Kubernetes.

  • @Brandon Tharp mentioned the notes on the wiki page: Deployment at TAMU.

  • There is not enough documentation to spin up all these containers so they have had to compile the source code.

  • Integration support needs to get on the roadmap.

  • @Tod Olson proposed that the project be consistent in the container organization tool used - should the project settle on one, Swarm, Kubernetes, etc., that would give us a better chance of moving forward together.

  • FOLIO is resource-intensive; need multiple docker containers. Vagrant doesn't work that way. Vagrant not suitable for production FOLIO environment.

  • Tod remarked that the group should find out where FOLIO is resource intensive through testing - can optimization be done?

  • Questions were raised about storage deployment, but that is unknown at this time. There's no routine multi-master model - rather difficult with Postgres. Unless you push "the magic button" to use Aurora with AWS.

    • What is the actual availability of the database - in the old model under Voyager a single Oracle database was sufficient for 20 years, with periodically scheduled downtimw.

    • Microservices are supposed to lend themselves to stateless architecture but that's hard to reconcile with traditional databases that are always "on."

  • Next meeting: LS tool system at Cornell. Demo with @Chris Manly and @Holly Mistlebauer


Action items