From: | Radoslav Nedyalkov <rnedyalkov(at)gmail(dot)com> |
---|---|
To: | Achilleas Mantzios <achill(at)matrix(dot)gatewaynet(dot)com> |
Cc: | pgsql-general(at)lists(dot)postgresql(dot)org |
Subject: | Re: Initial load from standby in logical replication. |
Date: | 2019-01-31 14:29:54 |
Message-ID: | CANhtRibf_MzMrmUUQPAEyT+ev8ryD2b55+NY3Zi78mcLk39f8Q@mail.gmail.com |
Views: | Raw Message | Whole Thread | Download mbox | Resend email |
Thread: | |
Lists: | pgsql-general |
On Thu, Jan 31, 2019 at 4:09 PM Achilleas Mantzios <
achill(at)matrix(dot)gatewaynet(dot)com> wrote:
> On 31/1/19 3:55 μ.μ., Radoslav Nedyalkov wrote:
> > Hi All,
> > We want to setup a logical replica with initial load from physical
> standby. (all pg10.5)
> > Basically what we tested is:
> > 1.on the master: create publication and logical slot.
> > 2.on the standby: pause the stanby. (lsn from 1. is already replayed).
> > 3.on the standby: get last replayed lsn, pg_last_wal_replay_lsn().
> > 4.export/import the data into the logical replica.
> > 5.on the logical replica: create subscription
> with pg_replication_origin_advance'd to the lsn from 3.
> > 6.on the logical replica: enable subscription.
> >
> > On a test setup with some bulk inserts results are encouraging.
> > Still we're not that confident. Does the procedure looks okay?
> > What if standby is paused at COMMIT lsn?
> Who is gonna consume data from the primary's replication slot? If there is
> no subscriber connected and reading data, then the primary will think that
> the subscriber is down and will accumulate WALs
> till some replication client is connected to it (which will not happen
> since the subscriber points to the standby).
>
Yes. that's clear. Primary will accumulate WALs until logical replica
connects to it.
The point is to offload the primary from massive initial copying.
> > Thanks!
> > Radoslav Nedyalkov
> >
> --
> Achilleas Mantzios
> IT DEV Lead
> IT DEPT
> Dynacom Tankers Mgmt
>
>
>
From | Date | Subject | |
---|---|---|---|
Next Message | Tom Lane | 2019-01-31 14:58:11 | Re: Old tsearch functions |
Previous Message | Achilleas Mantzios | 2019-01-31 14:08:33 | Re: Initial load from standby in logical replication. |