Server: Ubuntu server 14 lts + PostgreSQL 9.2 I want create cluster database using drbd, but i can't set PGDATA without cluster initialization. I just need say pgsql use data from drbd disk. How i can do it?
Example 1:
mkdir /cluster/var/lib/pgsql -p
chown postgres:postgres /cluster/var/lib/pgsql -R
cp -R /var/lib/pgsql /cluster/var/lib/pgsql
edit /etc/init.d/postgresql :
PGDATA=/cluster/var/lib/pgsql/data
...
PGLOG=/cluster/var/lib/pgsql/pgstartup.log
/etc/init.d/postgresql start
in postgresql 8.3 it works, but in 9.2 i can't change pgdata in /etc/init.d/postgresql, i need find another file and set pgdata, but, surprise, it's do nothing.
Example 2: PGDATA - Specifies the directory where the database cluster is to be stored; can be overridden using the -D option.
Ok, let's start: --pgdata=directory yeah, it's works! but now we have postgresql-xc and error like "postgresql don't know this user - postgresql". drbd start replicate data from cluster, but postgresql start it too.
UPD 1:
root: initdb --pgdata=/home/username/dir
~initdb not install~bla-bla-bla~use apt-get install postgres-xc
UPD2:
$: /usr/lib/postgresql/9.3/bin/initdb --pgdata=/whateveryouwant
#now you can run postgresql only one way:
$: /usr/lib/postgresql/9.3/bin/postgres -D /see_up
#then:
LOG: database system was shut down at 2014-09-26 15:56:33 YEKT
LOG: database system is ready to accept connections
LOG: autovacuum launcher started
#aaaaaaaaaaand...nothing. just empty console, ^C stopping postgres
#another SSH connect:
$: ps-ela
S 1000 5995 5217 0 80 0 - 62202 poll_s pts/0 00:00:00 postgres
1 S 1000 5997 5995 0 80 0 - 62202 poll_s ? 00:00:00 postgres
1 S 1000 5998 5995 0 80 0 - 62202 poll_s ? 00:00:00 postgres
1 S 1000 5999 5995 0 80 0 - 62202 poll_s ? 00:00:00 postgres
1 S 1000 6000 5995 0 80 0 - 62415 poll_s ? 00:00:00 postgres
1 S 1000 6001 5995 0 80 0 - 26121 poll_s ? 00:00:00 postgres
#is it ok? because...
$: /etc/init.d/postgresql status
9.3/main (port 5432): down
With Ubuntu, to use a specific data directory for a PostgreSQL instance, you just need to do:
# pg_createcluster --datadir=/path/to/the/directory 9.3 nameofcluster
You don't need or want to fiddle with $PGDATA
anywhere. See pg_createcluster manpage.
If using drdb
to replicate the data directory in a cluster, keep in mind that the term cluster
in PostgreSQL parlance has a different signification: it's always a cluster of databases, which means all the databases served by a specific PostgreSQL instance with its unique data directory. This differs from a drdb cluster
which is a cluster of machines.
The data directory of a PostgreSQL cluster cannot be shared at all, in the sense that only one specific instance of PostgreSQL running on a single machine may write into it, otherwise corruption will ensue.
If drdb
is configured to conform with this requirement, it's OK to use it, otherwise not.
You may also question why using drdb
in the first place when PostgreSQL has built-in replication options since version 8.3 that are more specialized and better suited to a database than pure disk replication.