Date: Thu, 25 Oct 2018 20:07:30 +0200
From: Johannes Truschnigg <johannes@truschnigg.info>
To: "Ghiurea, Isabella" <Isabella.Ghiurea@nrc-cnrc.gc.ca>
Cc: "pgsql-admin@postgresql.org" <pgsql-admin@postgresql.org>
Subject: Re: upgrade slave  in streaming replication from PG 10.4 to 10.5.1
Message-ID: <20181025180729.xoharwuuytno5b7j@vault.lan>
References: <1540479215718.82040@nrc-cnrc.gc.ca>
MIME-Version: 1.0
Content-Type: multipart/signed; micalg=pgp-sha512;
	protocol="application/pgp-signature"; boundary="b3542bj2is35tgcd"
Content-Disposition: inline
In-Reply-To: <1540479215718.82040@nrc-cnrc.gc.ca>
User-Agent: NeoMutt/20170113 (1.7.2)
Precedence: bulk


--b3542bj2is35tgcd
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
Content-Transfer-Encoding: quoted-printable

Hi Isabella,

On Thu, Oct 25, 2018 at 02:53:35PM +0000, Ghiurea, Isabella wrote:
> Gmorning List,
>=20
> we have a system running  PG streaming replication using repmgr4 , PG hos=
ts
> are running Cent OS 7 with  PG 10.4, we have a request to upgrade to PG 1=
0.5
> , my plan is to upgrade the slave db host first and later on /next month
> switch over and upgrade the existing PG master, we are using the following
> PG extensions: citext , pg_sphere,pg_stat_statements,pgrowlocks.

Your question reads to me as if you somehow missed the important change in
pg's version numbering scheme that 10.0 introduced: These days, 10.n to 10.=
n+1
is not considered a major release upgrade (of the sort that requires
pg_upgrade or dump+restore, etc.) any more; it's what (e.g.) 9.6.n to 9.6.n=
+1
used to be. Therefore, both versions you mentioned should be compatible and
allow physical replication between themselves, no matter if the master or t=
he
slave are on the newer (minor) version than the other.

> Also in existing PG 10.4 cfg:/data/ pg_wal is a link  and we have cfg
> multiple tablespaces

Shouldn't matter, if your existing configuration works. Just install the ne=
w,
upgraded binaries, restart your service(s), and you should be good to go.

Personally, I'd upgrade the slave first- if anything goes awry (very unlike=
ly,
but accidents happen!), you don't have to fail over that way.

> Are any issues   related to replication if we are upgrading the  PG slave=
 host first ,  ?

Not in this scenario - if you were doing an actual major release upgrade (1=
0.x
to 11.x or later), you'd have to get more creative and (probably - there are
options, if you cannot afford that kind of interruption, but they are a bit
more involved) break replication for a while. If it comes to that, I'm sure
someone the list will be able to help out.

Cheers, and best of luck with your upgrade procedure!

--=20
with best regards:
- Johannes Truschnigg ( johannes@truschnigg.info )

www:   https://johannes.truschnigg.info/
phone: +43 650 2 133337
xmpp:  johannes@truschnigg.info

Please do not bother me with HTML-email or attachments. Thank you.

--b3542bj2is35tgcd
Content-Type: application/pgp-signature; name="signature.asc"

-----BEGIN PGP SIGNATURE-----
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=gDph
-----END PGP SIGNATURE-----

--b3542bj2is35tgcd--