On 9/6/2016 8:21 PM, Venkata B Nagothi wrote:
Thats quite huge. pg_upgrade would be a better choice and yes, downtime is needed. You need to have the database shutdown all through the upgrade process.
How long it will take depends on type of Disks you have. It is definitely a lot faster than pg_dump/pg_restore.
Just moving a copy of the 2.3TB data between service providers is going to take quite a long time.
ok, just a rough concept.
1) setup new streaming slave at new provider using 9.2, wait for replication to complete and satisfy yourself that the database is complete and intact.
2) shut down master at old provider (and any other slaves), promote new provider 9.2 to master.
3) pg_upgrade using --link option to 9.4 on new provider. if this fails, restart old master and start over at step 1
4) bring up applications on new database. if they fail, restart old master, and start over at step 1
5) bring up new slave(s) on new database.
--
john r pierce, recycling bits in santa cruz
