|
Return Of JimmyJars posted:Of course that doesn't help when the colo customer below you comes in and rips all the cables out of your servers, removes them from the racks and just tosses them on the floor because "it was easier for me to work on my server." I can't stop reading this. I sort of want to photoshop it on top of a motivation scene like those demotivator posters. This the best thing I've read in this thread.
|
#
¿
Aug 15, 2013 01:51
|
|
|
|
| # ¿ May 7, 2024 21:39 |
|
|
incoherent posted:Your loving devs are lazy as all hell. This pattern of test servers becoming production is the green-blue pattern and is actually a great pattern. The idea is that you constantly deploy to test throughout preparing for a release, then clone the data out of production for the migration and switch the LB over to the previous test environment. You wait a few days in case of failures (failover is just changing the loadbalancer back to the previous cluster and syncing data). Once deployed for a while you then reinstall all the previous production machines and they become the test boxes for the next set of deployments. This works best if you're doing major releases at least once a month so the set of machines constantly cycle. It makes it very easy to stage and release major architectural changes with a safety net. Hell, it's even in the http://martinfowler.com/books/continuousDelivery.html book.
|
|
#
¿
Sep 7, 2013 14:48
|
|