I am managing a PostgreSQL database cluster composed by 3 nodes with Pacemaker + Corosync (in additional to normal PostgreSQL cluster facilities). I am working on a procedure to migrate this system from 9.4.9 TO 9.6.3. During the procedure I got the following error:
2017-09-13 10:46:29.014 GMTLOG: WAL file is from different database system: WAL file database system identifier is 6465195050841964088, pg_control database system identifier is 6465207123430201408`
Since the upgrade is on another data directory I realized that when I create a new empty database I didn't passed to the initdb the --sysid option to pass the old PostgreSQL system identifier. Since this message appear only when I startup master database with pacemaker (using crm -w node online ) and not when I start the normal postgres database, I imagine that pacemaker keep the postgres system identifier somewhere on filesystem. But I wasn't able to find it with a grep on the whole filesystem. I read pacemaker doc with no success. Can you help me to figure out where this information is kept (if it is kept) by pacemaker?
In addition to this I also found that after a while pacemaker realign this system identifier in someway because in some tests the problem does not appear, in other tests it simply fix by itself after a while.
So it's clear to me that in the migration I have to preserver the postgresql system identifier to avoid errors, but it's not clear to me how pacemaker manage it.
Thanks in advance for help.