Thread

1. Re: PITR, checkpoint, and local relations

   Vadim Mikheev <vmikheev@sectorbase.com> — 2002-08-02T22:01:11Z

   > > How do you get atomic block copies otherwise?
   > 
   > Eh?  The kernel does that for you, as long as you're reading the
   > same-size blocks that the backends are writing, no?
   
   Good point.
   
   Vadim
   
   

2. Re: PITR, checkpoint, and local relations

   J.R. Nield <jrnield@usol.com> — 2002-08-02T22:24:07Z

   Are you sure this is true for all ports? And if so, why would it be
   cheaper for the kernel to do it in its buffer manager, compared to us
   doing it in ours? This just seems bogus to rely on. Does anyone know
   what POSIX has to say about this? 
   
   On Fri, 2002-08-02 at 18:01, Mikheev, Vadim wrote:
   > > > How do you get atomic block copies otherwise?
   > > 
   > > Eh?  The kernel does that for you, as long as you're reading the
   > > same-size blocks that the backends are writing, no?
   > 
   > Good point.
   > 
   > Vadim
   > 
   -- 
   J. R. Nield
   jrnield@usol.com
   
   
   
   
   

3. Re: PITR, checkpoint, and local relations

   Richard Tucker <richt@multera.com> — 2002-08-07T15:32:01Z

   
   > -----Original Message-----
   > From: Mikheev, Vadim [mailto:vmikheev@SECTORBASE.COM]
   > Sent: Friday, August 02, 2002 6:01 PM
   > To: 'Tom Lane'; J. R. Nield
   > Cc: Richard Tucker; Bruce Momjian; PostgreSQL Hacker
   > Subject: RE: [HACKERS] PITR, checkpoint, and local relations
   >
   >
   > > > How do you get atomic block copies otherwise?
   > >
   > > Eh?  The kernel does that for you, as long as you're reading the
   > > same-size blocks that the backends are writing, no?
   >
   > Good point.
   >
   We know for sure the kernel does this?  I think this is a dubious
   assumption.
   > Vadim
   >
   
   
   

4. Re: PITR, checkpoint, and local relations

   Tom Lane <tgl@sss.pgh.pa.us> — 2002-08-07T15:40:34Z

   Richard Tucker <richt@multera.com> writes:
   >>> Eh?  The kernel does that for you, as long as you're reading the
   >>> same-size blocks that the backends are writing, no?
   
   > We know for sure the kernel does this?  I think this is a dubious
   > assumption.
   
   Yeah, as someone pointed out later, it doesn't work if the kernel's
   internal buffer size is smaller than our BLCKSZ.  So we do still need
   the page images in WAL --- that protection against non-atomic writes
   at the hardware level should serve for this problem too.
   
   			regards, tom lane