2019-03-08 - System Operations and Management SIG Agenda and Notes

• Welcome new member: Philip Robinson
• Introduction round
  
  • Patty and Todd work together for UA go-live
  • Cornell plans to go live June 2020 (at the end of the FY).
• Find a note taker: Ingolf

Meeting Notes:

Jason presents a draft of a write-up of the deployment issues that were brought up (mainly by the TAMU team) during the last few sessions. Jason will put the draft into the SysOps Google Drive folder and keep working on it.

Jason needs to intergrate several etcd nodes (I found this site which explains etcd: What is etcd and how to set-up an etcd cluster).

We need to move dependency resolution to a different tool (than Okapi). Right now, you have to have Okapi run before you start up a module.

Q (Robbie): Does this include having to deploy a single module within an existing system ?

A: A new module might depend on new modules with new interfaces. Okapi spits out a list and you have to re-deploy all those modules. There is no easy upgrade path without using Okapi at the moment. But that's the point to use an orchestration tool, not having to deal with all the dependent modules.

Database changes involved in the mix are also an issue. Right now, everyone has to hand-upgrade the database, has to "hand-massage" the database. That's an issue of top priority. If you will have a shorter release cycle than quarterly, it's impossible to keep up.

(Stephen) Third, there needs to be a convention for rolling back. How do we roll modules forward and backward ? It's a nightmare scenario having to cleanup data. The system should handle the DB changes that need to be rolled back.

(Anton) Okapi should pick up a list of modules and deploy it, not the other way round. We need issues, stories (in JIRA) and a Backlog.
Anton offers to work on the JIRA issues for SysOps, to put the issues into JIRA. Anton will meet with Adam Dickmeiss and Jason to work on the features and stories. Adam was chiefly involved in the development of Okapi.

Also upgrades (of modules) are extremely important.

(Anton) Chalmers will go live this summer. By Q4 (→ Q1 ? IK) we should have a path to all of this. These things need to be added before anyone can go live. Chalmers doesn't use the community version (of what ) , so things can not be re-used by us. Chalmers are using AmazonWS, probably with AKS.

(Anton) The core dev team has recently split up into core-functinal and core-platform. They have also just started using a SCRUM process. A new EPAM team has been formed, they are "getting wet" with SCRUM. SCRUM is about usability and accountability. SCRUM is a procedure model in agile development.

(Jason) In contrast to DevOps, we don't have a lot of interactions with the developers in this group.