From: | Francisco Olarte <folarte(at)peoplecall(dot)com> |
---|---|
To: | Jeff Baldwin <tarheeljeff(at)gmail(dot)com> |
Cc: | Alan Hodgson <ahodgson(at)lists(dot)simkin(dot)ca>, "pgsql-general(at)postgresql(dot)org" <pgsql-general(at)postgresql(dot)org> |
Subject: | Re: Migrate 2 DB's - v8.3 |
Date: | 2016-05-28 15:26:27 |
Message-ID: | CA+bJJbwpp7Zr+TYWxg2iv07pdcY+T_1Smi31Lr0p4GKk1f-sAg@mail.gmail.com |
Views: | Raw Message | Whole Thread | Download mbox | Resend email |
Thread: | |
Lists: | pgsql-general |
Jeff:
On Sat, May 28, 2016 at 12:38 AM, Jeff Baldwin <tarheeljeff(at)gmail(dot)com> wrote:
> Thank you for your time Alan.
..
> To move the DB, you are suggesting something like this:
> pg_dump -h dbms11 -U postgres -C mls11 | psql -h localhost -d mls11 -U
> postgres
I'd like to point one thing, you MAY get a little more speed if you
run pg_dump AND psql each in the same host as the DB it's operating on
to minimize latency ( and I would time unix socket vs network first in
case it differs ). ( to do that I would try something like 'ssh dbms11
"pg_dump mls11 " | psql -d mls11' with all the needed doodahs, and
maybe use something like netcat or socat instead of ssh ). The
rationale being the intermediate dump is just a data stream and not
latency sensitive ( except for the window*latency problem, but you are
not going to hit that on a LAN ), while the dump/restore does DB work
which is more latency sensitive ( I do not know how many RTTs it would
need, specially with blobs, but you can try it ).
¿ How many hours does it take in your tests? Because if you have 1-2
and you can do the dump psql pipe trick, which is quite robust, in 3-4
you may push for it ( arguing it's a simpler an more testable process
).
Francisco Olarte.
From | Date | Subject | |
---|---|---|---|
Next Message | Jeff Janes | 2016-05-28 17:32:15 | Re: swarm of processes in BIND state? |
Previous Message | Martín Marqués | 2016-05-28 14:38:56 | Re: BDR to ignore table exists error |