Kendra Little gives thoughts on getting to zero-downtime releases:
I’ve worked in a bunch of environments where we rolled schema changes to production throughout the week, and worked hard to limit downtime. Our biggest challenges were:
-
The code released wouldn’t have the intended effect, and multiple hotfixes would sometimes need to be written quickly if the issue was urgent.
-
Problems would happen with replication in production – either replication would be delayed because a lot of modifications were occurring, or a change in an “upstream” system would cause a huge amount of data processing to happen in a “downstream” database
I’m in a ZDT environment. There’s a pretty good amount of overhead work that goes into getting there; there are times in which a 20 minute window would have saved hours of development time, so it’s important to keep that trade-off in mind.