On 9/18/19 8:31 PM, David Steele wrote:
> On 9/18/19 6:59 PM, Ron wrote:
>> Scenario: there's data corruption on production server, so we need to do
>> a PITR restore from "a few days ago" of the cluster holding the prod
>> databases to a second cluster on that same VM in order to try and find
>> the missing data and load it back into the prod cluster.
>>
>> Other than putting a high I/O load on the LUN where repo-path is located
>> (from both writing WALs to it and reading the backed up files), will
>> there be any problems when "pg_ctl start" processes recovery.conf and
>> applies the WAL files to the new cluster while the prod cluster is
>> writing new WAL files.
>>
>> Does my question make sense?
> It does, but the answer lies outside of pgBackRest. "Can the repo
> storage handle the load of archive-push and archive-get at the same
> time" is really a question of storage and network throughput.
That's outside my control and will "just" slow things down.
> pgBackRest compresses everything by default which goes a long way
> towards increasing throughput, but ultimately we don't control the
> bandwidth.
>
> Having said that, if the storage and network throughput are sufficient,
> restoring and recovering a standby using pgBackRest will not impact the
> primary as a direct pg_basebackup will.
I'm concerned with one pgbackrest process stepping over another one and the
restore (or the "pg_ctl start" recovery phase) accidentally corrupting the
production database by writing WAL files to the original cluster.
--
Angular momentum makes the world go 'round.
