Thread: Don't include MMAP DSM's transient files in basebackup
The files are not useful when restoring a backup and would be automatically deleted on startup anyway. Also deleted an out-of-date comment in dsm.c. / Oskari
Attachment
On Wed, Jul 6, 2016 at 10:46 PM, Oskari Saarenmaa <os@aiven.io> wrote: > The files are not useful when restoring a backup and would be automatically > deleted on startup anyway. The same could be said about pg_snapshots. -- Michael
Hi, On 2016-07-06 16:46:53 +0300, Oskari Saarenmaa wrote: > The files are not useful when restoring a backup and would be automatically > deleted on startup anyway. Also deleted an out-of-date comment in dsm.c. No arguments against the change. But I am wondering how you ran into this - the mmap dsm backend really should pretty much never be used... - Andres
On Thursday, July 7, 2016, Andres Freund <andres@anarazel.de> wrote:
--
Magnus Hagander
Me: http://www.hagander.net/
Work: http://www.redpill-linpro.com/
On 2016-07-06 16:46:53 +0300, Oskari Saarenmaa wrote:
> The files are not useful when restoring a backup and would be automatically
> deleted on startup anyway. Also deleted an out-of-date comment in dsm.c.
No arguments against the change. But I am wondering how you ran into
this - the mmap dsm backend really should pretty much never be used...
Unfortunately I think that's still what happens by default if you use pg_upgradecluster on debian :(
//Magnus
--
Magnus Hagander
Me: http://www.hagander.net/
Work: http://www.redpill-linpro.com/
On 2016-07-07 08:49:18 +0100, Magnus Hagander wrote: > On Thursday, July 7, 2016, Andres Freund <andres@anarazel.de> wrote: > > > On 2016-07-06 16:46:53 +0300, Oskari Saarenmaa wrote: > > > The files are not useful when restoring a backup and would be > > automatically > > > deleted on startup anyway. Also deleted an out-of-date comment in dsm.c. > > > > No arguments against the change. But I am wondering how you ran into > > this - the mmap dsm backend really should pretty much never be used... > > > > > Unfortunately I think that's still what happens by default if you use > pg_upgradecluster on debian :( Ugh. Sounds like that should better be fixed before 9.6. It'll make parallel query seriously slower. Andres