ansible playbok to maintain remote servers

[Karl Auer]

On Mon, 2023-01-09 at 13:58 +0100, robert rottermann wrote:
> However I wonder who this is done when you have to maintain hundreds
> of servers.

Yeah... 20 is sort of a difficult number. Too few to bother with full
automation, too many to do manually. If you have hundreds you DO
automate it, but you also have ways to test it, and ways to roll
everything back if it all goes wrong. It depends a lot on whether the
systems are pets or cattle - pets are unique snowflakes, each one
different and important. Cattle are interchangeable, reproducible,
disposable and recreatable. It also depends on what the systems are
doing, and whether they are doing it redundantly, scalably etc, and
what performance impact is acceptable, either for the update process or
for the recovery process.

A concrete example:

In my job, among other things we run highly-available, scalable
WordPress webservers.

To update these, we build new server images (minus the web data) and
test them outside the production systems. The build of the updated and
upgraded image is automated; the testing is manual. Then we drop the
tested image into the scaling mechanism and the rest is automatic. The
scaling mechanism replaces every instance by killing the old ones as
new ones are launched, so the right number of instances are always
running. We avoid failure by pre-testing; but if something goes wrong
even so, we just put the old image back into the pipeline.

Regards, K.

PS: Lots of people think they know how to make WordPress scalable andavailable, but they are almost always wrong :-)
 
-- 
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Karl Auer (kauer at biplane.com.au)
http://www.biplane.com.au/kauer