Leeds DevOps Sep 2019
We had an excellent DevOps meeting tonight, with James Healey from Carbon DMP, and Matthew Skelton from Conflux
presenting.
James explained how he and his team did monitoring, alerting, and support. (Side note: he’d presented before at a
prison as part of charitable education work there,
https://www.code4000.org I think). As of recently, Azure and Rancher have good
monitoring capabilities natively. Azure for stock or custom queries, and Rancher for CPU utilisation etc. Azure and
AWS cost explorers have also recently had developments to make the costing a lot more visualisable. They use lots of
familiar tooling; AWS, Rancher, Pager Duty, Trello, Kubernetes.
Other things to check out mentioned in the talk were Certcheckr and Kube-backup.
Matthew talked about the context and succession of The Spotify Model (
there is no Spotify model). From a number of
countries, businesses, and studies, some findings and principles arose. The talk arrived at some of these and more
depth can be found in the book which is out
tomorrow.
Ideas included:
Team first thinking – considering teams as the production units and fitting work to them
Choosing team boundaries carefully
Considerations for the physical and digital workspaces; the Reverse Conway Effect which is to preemptively order teams
as per the interactions expected of an efficient system
There are 3 fundamental team interaction modes: Collaboration, Provider/Consumer, and Facilitator/Follower.
Read later – Conway’s 1968 paper
Link: https://www.meetup.com/LeedsDevops/events/264672882/