Re: NetBSD on a database server - WAPBL, softdeps

[Steven Bellovin <smb%cs.columbia.edu@localhost>]

On Sep 17, 2009, at 10:54 PM, Thor Lancelot Simon wrote:

> On Fri, Sep 18, 2009 at 01:21:06AM +0100, Mindaugas Rasiukevicius  
> wrote:
> > Thor Lancelot Simon <tls%panix.com@localhost> wrote:
> > > > and keeping in mind softdeps are scheduled to be removed in 6.0
> > > > (http://mail-index.netbsd.org/netbsd-announce/2008/12/14/msg000051.html 
> > > > ),
> > > > what is the recommended configuration, filesystem wise, for a  
> > > > database
> > > > server running on NetBSD? Wouldn't running with "bare" UFS be too  
> > > > slow?
> > >
> > > Why would you expect any benefit from either WAPBL or soft updates  
> > > on
> > > a database server?  Databases (or, at least, any database worth  
> > > using)
> > > manage their tablespaces independently, generally without creating  
> > > or
> > > deleting files in the filesystem -- or doing so very seldom, if at  
> > > all.
> >
> > Are you suggesting that long boots after system failure is the  
> > option?
>
> I'm suggesting that long boots after system failure *simply won't  
> happen*
> with sane disk layout and database design.  Even an FFS filesystem of
> many terabytes will fsck very quickly if it contains only the very  
> small
> number of directories and files that a properly designed database  
> uses.
>
> Databases journal their data themselves.  They don't need help from  
> the
> filesystem to do it, generally speaking.
Mmm -- maybe.  I think they need some way to ensure that certain  
blocks are
flushed to the disk.  This is, strictly speaking, not a filesystem  
issue, but
something has to pass the request (from fsync_range()?) to the  
driver.  I note
that the MacOS F_FULLFSYNC command of fcntl() is only specified to  
exist for
a few file systems, at least on Leopard.

Other than that, I agree with your points completely.

                --Steve Bellovin, http://www.cs.columbia.edu/~smb