Обсуждение: postmaster/postgres options assimilation plan
Here's the plan for assimilating the command-line options of the postmaster and postgres options. I reported earlier on a couple of conflict areas; here is the full plan: * Remove: postmaster -a -b -m -M These options have done nothing forever. * postmaster options added to postgres: -h -i -k -l -n These options will not have any useful effects, but their behavior is consistent if you do, say, SHOW listen_addresses. * postgres options added to postmaster: -e -E -f -O -P -t -W Using -E with postmaster does nothing, though. * Renamed options (because of conflicts): postgres -N is now postgres -j (mostly internal use) postgres -o is now postgres -r (mostly internal use) (postmaster -o is obsolete but still works for compatibility; postgres -o will get you an error.) postgres -p is now postgres -y (internal use only) postmaster -S now sets work_mem, like postgres -S does. The (deprecated) silent mode can be obtained using the long option --silent-mode=on. postmaster -s is now postmaster -T (expert/hacker use only) For the options -O, -P, and -W I have added three new GUC variables allow_system_table_mods (PGC_SIGHUP), ignore_system_indexes (PGC_BACKEND), connection_startup_delay (PGC_BACKEND); mostly to simplify the postmaster->postgres communication. Except for a few odd exceptions, all command line arguments now map to setting a GUC variable. Comments?
Peter Eisentraut <peter_e@gmx.net> writes: > Here's the plan for assimilating the command-line options of the postmaster > and postgres options. > ... > * postmaster options added to postgres: -h -i -k -l -n > These options will not have any useful effects, but their behavior is > consistent if you do, say, SHOW listen_addresses. Did you mean -N rather than -n here? > For the options -O, -P, and -W I have added three new GUC variables > allow_system_table_mods (PGC_SIGHUP), ignore_system_indexes (PGC_BACKEND), > connection_startup_delay (PGC_BACKEND); mostly to simplify the > postmaster->postgres communication. Might as well let allow_system_table_mods be PGC_BACKEND too. The dangers it poses are not made less by turning it on in all backends at the same time. regards, tom lane
FYI, with the options merged, we still have this TODO item: * %Remove behavior of postmaster -o --------------------------------------------------------------------------- Peter Eisentraut wrote: > Here's the plan for assimilating the command-line options of the postmaster > and postgres options. I reported earlier on a couple of conflict areas; here > is the full plan: > > * Remove: postmaster -a -b -m -M > > These options have done nothing forever. > > * postmaster options added to postgres: -h -i -k -l -n > > These options will not have any useful effects, but their behavior is > consistent if you do, say, SHOW listen_addresses. > > * postgres options added to postmaster: -e -E -f -O -P -t -W > > Using -E with postmaster does nothing, though. > > * Renamed options (because of conflicts): > > postgres -N is now postgres -j (mostly internal use) > > postgres -o is now postgres -r (mostly internal use) > > (postmaster -o is obsolete but still works for compatibility; postgres -o will > get you an error.) > > postgres -p is now postgres -y (internal use only) > > postmaster -S now sets work_mem, like postgres -S does. The (deprecated) > silent mode can be obtained using the long option --silent-mode=on. > > postmaster -s is now postmaster -T (expert/hacker use only) > > > For the options -O, -P, and -W I have added three new GUC variables > allow_system_table_mods (PGC_SIGHUP), ignore_system_indexes (PGC_BACKEND), > connection_startup_delay (PGC_BACKEND); mostly to simplify the > postmaster->postgres communication. > > Except for a few odd exceptions, all command line arguments now map to setting > a GUC variable. > > Comments? > > ---------------------------(end of broadcast)--------------------------- > TIP 2: Don't 'kill -9' the postmaster > -- Bruce Momjian | http://candle.pha.pa.us pgman@candle.pha.pa.us | (610) 359-1001+ If your life is a hard drive, | 13 Roberts Road + Christ can be your backup. | Newtown Square, Pennsylvania19073