Horton

Re: Refactoring the checkpointer's fsync request queue

Andres Freund <andres@anarazel.de>

View thread View on postgr.es
From: Andres Freund <andres@anarazel.de>
To: Shawn Debnath <sdn@amazon.com>
Cc: Thomas Munro <thomas.munro@gmail.com>, Robert Haas <robertmhaas@gmail.com>, Tom Lane <tgl@sss.pgh.pa.us>, Dmitry Dolgov <9erthalion6@gmail.com>, PostgreSQL-development <pgsql-hackers@postgresql.org>
Date: 2019-04-04T04:35:59Z
Lists: pgsql-hackers 
On 2019-04-03 21:19:45 -0700, Shawn Debnath wrote:
> On Thu, Apr 04, 2019 at 02:01:14PM +1300, Thomas Munro wrote:
> > On Thu, Apr 4, 2019 at 11:39 AM Thomas Munro <thomas.munro@gmail.com> wrote:
> > > ... Perhaps
> > > that is an argument for putting the sync handler number *inside* the
> > > FileTag, since we currently intend to do that with smgr IDs in
> > > BufferTag (stealing space from ForkNumber).
> > 
> > Here is a version like that.  I like it better this way, and the extra
> > space can be clawed back by using 16 bit types to hold the fork number
> > and sync handler number.
> 
> +typedef struct FileTag
> +{
> +	int16		handler;		/* SyncRequstHandler value, saving space */
> +	int16		forknum;		/* ForkNumber, saving space */
> +	RelFileNode rnode;
> +	BlockNumber segno;
> +} FileTag;

Seems odd to me to use BlockNumber for segno.

Commits

  1. Fix bugs in mdsyncfiletag().

  2. Refactor the fsync queue for wider use.

  3. Don't forget about failed fsync() requests.

  4. PANIC on fsync() failure.

  5. Move LockClauseStrength, LockWaitPolicy into new file nodes/lockoptions.h.

  6. Add new file for checkpointer.c

  7. Split work of bgwriter between 2 processes: bgwriter and checkpointer.

  8. Install infrastructure for shared-memory free space map. Doesn't actually