From: | Jean-Christophe Arnu <jcarnu(at)gmail(dot)com> |
---|---|
To: | Nicolas Lutic <n(dot)lutic(at)loxodata(dot)com> |
Cc: | pgsql-hackers(at)lists(dot)postgresql(dot)org |
Subject: | Re: PITR on DROP DATABASE, deleting of the database directory despite the recovery_target_time set before. |
Date: | 2019-12-13 09:09:53 |
Message-ID: | CAHZmTm1X6OhB3dQDw0vyc27Gb3WeO0g3a0YSEPThPxY_-FahEg@mail.gmail.com |
Views: | Raw Message | Whole Thread | Download mbox | Resend email |
Thread: | |
Lists: | pgsql-hackers |
Hello,
Le mar. 19 nov. 2019 à 16:15, Nicolas Lutic <n(dot)lutic(at)loxodata(dot)com> a écrit :
>
> We are aware that this part is tricky and will have little effects on
> normal operations, as best practices are to use xid_target or lsn_target.
>
> I'm working with Nicolas and we made some further testing. If we use xid
target with inclusive to false at the next xid after the insert, we end up
with the same DELETE/DROP directory behaviour which is quite confusing. One
have to choose the xid-1 value with inclusive behaviour to lake it work.
I assume this is the right first thing to document the behaviour. And give
some examples on this.
Maybe we could add some documentation in the xlog explanation and a warning
in the recovery_target_time and xid in guc doc ?
If there are better places in the docs let us know.
Thanks
--
Jean-Christophe Arnu
From | Date | Subject | |
---|---|---|---|
Next Message | Amit Kapila | 2019-12-13 09:16:20 | Re: PATCH: logical_work_mem and logical streaming of large in-progress transactions |
Previous Message | Magnus Hagander | 2019-12-13 08:56:49 | Re: non-exclusive backup cleanup is mildly broken |