3 ms·
I've tried to setup a replicated postgres with autofailover and it honestly is a pita. the only sources of failover are rando scripts over the internet, that y
by LoSboccacc 10y ago
I've tried to setup a replicated postgres with autofailover and it honestly is a pita.
the only sources of failover are rando scripts over the internet, that you have to download hammer in to your version dialect and hope you don't trigger one of the many uncovered failover modes.
sure log shipping works, but that's far, FAR from a working solution. the gap requires ton of development hour, testing etc.
can't really blame people for using things with mature tooling.
(but I didn't switch to MySQL that'd be madness! there are plenty of good, replicated redundant stores out there, both nosql and sql)
- fanf2 10y agoDoes repmgr handle most of this?
- LoSboccacc 10y agomost of it. but in the master-slave configuration the clients can only work if connecting to the master, and when the master switch, scripts needs to go to each clients and update their config. not immensely bad if you also use pgbouncer so you can do it on the fly without restarting the whole client, but exceptionally vulnerable to split brains and the like.
- snuxoll 10y agoI've handled this with a DNS CNAME record that points to the active master, though pgBouncer is a valid solution as well.