2020-01-31 - System Operations and Management SIG Agenda and Notes

• Q: do we want the video recordings of our SysOps Meeting at WolfCon uploaded to our SIG space ? => YES

6 month ago Kubernetes subgroup was launched

Production install in a Kubernetes environment

Jason, Taras Spachenko et.al. contributed. There was no unified approach at first. The charge was to discuss and document Kubernetes deployment.

Changes to Okapi to make Kubernetes integration easier. They met regularly for 4 month. Then they tapered off their work. They want to go back of the SysOps group. Most libraries are not looking at hosting theirselves. Texas A&M is an excpetion in that. Most want to host it at some hosting provider.

Maybe they started early. Looking at features in folio-testing. There is going to be a preview-environment that PO's are going to use. Some features should not go to Master. Folio-realted development to folio integration. Prioritizing of tickets. Not many tickets have been prioritozed. Ability for Okap to use a Wildcard-<something>. That ticket had been prioritized.

Okapi needs to be made aware of interface changes. There hasn't been a lot of development to smooth that out. JSON module descriptors. Ian Hardy put out that tooling. Using ANSIBLE as config management tool. Managing dev ops infrastructure management tool. Treating Okapi as configuration management. Running scripts to keep Okapi in the proper state. Jason: We are not at that point that. Each module should have a distinguished format of configuration (Jo). Wayne: Yes.

Best Practices: In the SysOps SIG Google Folder, there is a Kubernetes soubgroup Folder. It contains recordings, some diagrams and a SysOps guide. That is our contribution to the community. Diagram: EKS Folio Cluster Network. Running Folio on EKS.2nd Diagram: Kubernetes FOLIO namespaces. Kubernetes Logical Architecture. Namespaces could be "staging areas" or "production areas".

More documentation: In the folio-install repository. There is a kube-rancher branch. (So far on Daisy, but will be updated). subfolder TAMU. This display the full infrastructure (16 steps).

SysOps Guide Work plan.

Ian Walls: OK, I have a Kubernetes Cluster. How do I make this work in FOLIO. I want a documentation and guideline for that.

Wayne: Want to have a walk-through with minikube.

• Short summary
• What have we learned ?
• Focus points of the discussions
• What concerns and suggestions were raised ?
• What did we accomplish well ? Kudos to the group for the work done in 2019 by Ingolf.
• What do we have to do now ?
• What's next

Update strategies

database management stuff

some new faces

some people heard this stuff the first time

... this discussion centers around dependency management → the tech debt list from the TC

call what we have now "V.1" of the FOLIO architecture - need to talk about what V.2 will look like

dependencies - maybe have to recombine some modules

a kind of messanging system for interaction between the modules. We need more understanding of how modules communicate with each other.

database strategies: At A&M: crunchy postgres, Amazon people: RDS, Chris: pgpool

upgrade strategies: moving data between releases  ; in Q4 upgrades will be more an extra transform and load, rather than an automated upgrade

other topic: performance as we start to onboard bigger libraries

there is no big database set to test with; Wayne: UChicago has released a large inventory data set. They were testing individual components and not entire workflows.

Need to simulate things like "30 simultaneous user". This kind of performance testing we need to do. Need to get that running as soon as we can.

The PubSub has been released after Q4. The messaging architecture ; Kafka.