Re:It is done, the PostgreSQL community rocks
by arjie on Wednesday April 05, 2017 @03:21PM
(#14044768)
Thanks for the advice. Is there a sane setup guide here with sane defaults for the master-slave replication?
I would rather not discover afterwards that I was supposed to `ulimit` the postgres process to some CPU in order to be able to successfully kill queries that are jamming the box.
My long answer: After years of experience with highly-available database provision with PostgreSQL, I concluded that I don't want automatic failover. It's more practical to keep a master running and constantly replicated by slaves that can be promoted to master in migration or failure situation. However, when there is a failure situation it's so exceptional that someone really needs to understand what happened before allowing the next victim (er, database) to go on stage.