We tried in production, pg_upgrade works well. But running vacuumdb , resulted in huge spike in CPU, system halted. Is there a way to fasten or parallel vacuum solution for faster recovery after pg_upgrade.
Our database size is around 500GB, contains multiple databases, huge records. What is the minimum way to do a vacuuming after pg_upgrade? This is for migration from 9.3 to 9.4.
Thanks, i resolved by following other forums. seems we should not have tablespace inside data ( even symlink) as well. Once i moved the tables space to other location & updated the symlinks, pg_upgrade completed very fast within 4 mins. Thanks a lot for your support.
I created CentoS, moved the data of 9.3 , completed the setup. later installed 9.4 , I could see pg_upgrade in that.
We have two tables space, which are symlink to directory just before data directory
#ls -lrt
/data/pgsql93/<tb1>
/data/pgsql93/<tb2>
/data/pgsql93/data/ --> here is the 9.3 pgdata presents
Inside /data/pgsql93/data/pg_tblspc/ , two folders are symlined into above folders(tb1 & tb2)
While doing pg_upgrade with link option, Do i need to copy the table space (tb1 & tb2) to the new pgsql94 directories? or Create Folders structure similar to pgsql93 tablespace?
i got an error while doing pg_upgrade with link option
psql:pg_upgrade_dump_globals.sql:91: ERROR: directory "/data/pgsql93/<tb1>/PG_9.4_201409291" already in use as a tablespace
I have installed postgres-contrib package as well. But pg_upgrade is not available. From this forum, https://forums.aws.amazon.com/thread.jspa?threadID=227761 amazon didnt provide pg_upgrade, thats the concern. Am going to try out with CentOS with existing DB Data, and use RDS import using migration service.
