On 9/22/14, frantisek holop <min...@obiit.org> wrote: > > i installed postgresql and starting to use it for > first time. here is what i think could be improved.
Hmm... > 1. > i think the postgres package readme could mention, > that while there is no default database location > the rc.d script has /var/postgresql/data hardcoded > in it. > > > 2. > without actually initializing a database i tried: Why did you do this? The README is clear about creating the database: "If you are installing PostgreSQL for the first time, you have to create a default database first." > $ sudo /etc/rc.d/postgresql start > postgresql(ok) # ??? > $ sudo /etc/rc.d/postgresql check > postgresql(failed) > > a bit of "set -x" reveals: > $ sudo su -l -c daemon -s /bin/sh _postgresql -c \ > "/usr/local/bin/pg_ctl -D /var/postgresql/data \ > start -l /var/postgresql/logfile" > server starting > $ echo $? > 0 > # tail -1 /var/postgresql/logfile > postgres cannot access the server configuration file > "/var/postgresql/data/postgresql.conf": No such file or directory > > what happens is, that pg_ctl's return code cannot > be trusted because it starts the server in background > and does not wait for confirmation. to really make > sure the server was started, -w must be used: > > $ sudo su -l -c daemon -s /bin/sh _postgresql -c \ > "/usr/local/bin/pg_ctl -w -D /var/postgresql/data \ > start -l /var/postgresql/logfile" > waiting for server to start........ stopped waiting > pg_ctl: could not start server > Examine the log output. > $ echo $? > 1 > > i dont know how long it takes to start up a huge > postgres installation with -w, and maybe it is good > to have it start in the background at startup time, > but it can mask startup errors like this... > > > 3. > my last issue is "check": i think it is a bit of overkill > that i have to be root to check if postgres is running. > this is because daemon_user is defined. so not even > '_postgresql' is allowed to check, although it > is allowed to run the actual 'pg_ctl check' command. > > i think it would make sense to get rid of rc_usercheck=NO, > all of rc_check() and override pexp= to the actual > daemon signature instead of pg_ctl.. this way any user > can run the check. The database could be in a shutdown mode, and the change you propose could give a false-positive. Maybe there are reasons why folks who have maintained this port, and have used it in production, for much longer than "first time", have things this way. --patrick > please find attached my patch that tries to remedy > the issues mentioned above. > > -f > -- > the best way out of a difficulty is through it. >
