Hello everyone,
Looking for some guidance: I'm trying to run 9.3.4 on Ubuntu 12.04 and 14.04, and everything seems okay for the most part, except the server won't start after reboot.
I believe the problem stems from the init.d script exiting before executing the start function. More specifically, a catch 22 emerges because the call to pg_lsclusters before starting fails because the socket dir doesn't exist, but the start action is supposed to take care of creating the socket dir if it doesn't exist, but we don't make it to the start function because the call to pg_lsclusters fails.
Because /run is tmpfs, it seems like the init script should never work after a reboot, which clearly can't be what other people are experiencing or I'd think there would be an uproar.
I've installed 9.3 on 2 machines that previously had 9.2 on them (one ubuntu 12.04, one 14.04) and I've installed 9.3 on a clean machine (14.04). All three seem to be affected by this weirdness.
If I take the logic from the start function that creates the socket dir and place it before the call to pg_lsclusters in the init script, the server starts without a problem.
The version of postgresql-common I'm running is 154.pgdg12.4+1 on the 12.04 machine and 154.pgdg14.04+1 on the 14.04 machines.
Has anyone ever run into something like this? Am I screwing something up?
