On 11/28/2016 10:08 PM, Stephen Davies wrote:
My databases were off the air for 24 hours.
On the up side, you've learned a couple of important lessons about
operating production services. The first of them is this: Set up a test
environment and actually test your changes there, first.
Increasingly, the other best practice is this: don't upgrade your
systems in-place. Decouple your data and your software. When you need
to update your software, build a new system and connect your data to
it. Automate the process. With automation in place, this process is
easier than upgrading your services, and provides a fall-back mechanism
in the event of failure. I was reluctant to adopt that mode of
operation, at first, but as I get closer to that practice, I find
operations less laborious and more reliable.