What is the best practice either reference value in terms of an SLA spec to expect the infrastructure to:
Sure you can say "how important is that" but somewhere it gets irrational. SLA offers rational approach.
Service Level Agreement
https://en.wikipedia.org/wiki/Service-level_agreement
https://www.atlassian.com/it-unplugged/best-practices-and-trends/stop-hating-on-slas
https://www.cio.com/article/2438284/outsourcing/outsourcing-sla-definitions-and-solutions.html
What will happen if the docker registry is down? Are all departments blocked? What is the impact?
If all production images reside in a docker registry and someone wants to release a newer version to production, but the registry is down, is it still possible to release the product?
SLA is about defining processes, checking the impact of each of them and the more important the higher the SLA.
One could this tool to choose a certain SLA for a certain procedure:
A SLA of 99.999 https://uptime.is/99.999 means a yearly downtime of 5m 15.6s, while https://uptime.is/99.9 means 8h 45m 57.0s.
One could set a SLA of 99.999 for the docker processes, but is a downtime of 5m 15.6s really required or is 8h 45m 57.0s in a year acceptable? If one hosts an ecommerce website and the webshop is down a couple of hours this will cause a lot of harm for the company as customers could not buy products and will probably buy products somewhere else.
For the docker processes that are used by developers and for releases to production one could argue that the first one could be 99.9 and the latter 99.999.
