ZeroDowntime_Abstract.txt

DevOps Flow Series, part 4 of 4

Title:
Zero downtime database deployments

Abstract:
You understand that big bang deployments are painful, but 
each deploy may disrupt users, so you are restricted to 
tight deployment windows for your database. The consequences 
of missing your windows aren't worth thinking about - but it 
gets more difficult each time.

In order to move to smaller, more frequent and more reliable 
deployments you need to demonstrate that you can safely 
deploy during business hours - without disrupting users.

In this session we'll discuss the most popular strategies: 
Blue/Green and Canary deploys, Cluster Immune Systems and 
Feature Toggles. Then we'll discuss the fundamental changes 
that we need to make to the way we think about database 
"deployments" and "releases" in order to support these 
strategies.

We will finish with a SQL Server demo that uses a feature
flag and the expand/contract model to support backward
compatibility and safe, fast rollback options.

Short description
Moving to a safer release cadence, based on smaller, more 
frequent deployments, normally requires deployments to occur 
while the systems are running. That’s scary. In this session 
we’ll discuss how to do it safely.

More from the DevOps Flow series:
1. DevOps 101 for Data Folks:
   https://gist.github.com/Alex-Yates/efef91357c6bd0dc1fbe63f2eb464f68
2. Simple database deployments with GitHub, Azure DevOps Services and SSDT: 
   https://gist.github.com/Alex-Yates/43c7f4d51a651cd8b3dcdcb48843691f
3. Solving the dev database problem with GitHub, Docker and dbaclone:
   https://gist.github.com/Alex-Yates/8125be587e58863ad35e0845ab6ffd5a
4. Zero downtime database deployments
   https://gist.github.com/Alex-Yates/ac4c791bbd98eb3269f35db59cc30f88