[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]

Re: ext3 performance issue with a Berkeley db application



Hi,

On Mon, 2003-02-03 at 17:51, Andreas Dilger wrote:
> On Feb 03, 2003  12:08 +0100, Matthias Andree wrote:
> > Are you aware of a module that applies to recent kernel versions and
> > that traces block numbers of ll_rw_block()? It might turn up some useful
> > information -- then we'd easier know the ext3 "output" to the hard disk;
> > we already know the "input" from the application at the syscall level.
> 
> Stephen Tweedie has a loopback driver:
> 
> 	 http://people.redhat.com/sct/patches/testdrive/
> 
> which allows you to trace all I/O.

And for debugging specific problems, it ought to be easy to extend that
to add (a) the buffer type (ie. data, metadata, journal IO) to the
tracing, plus the caller responsible for doing the IO.  I suspect here
that we may be getting interactions between the ext3 commit logic trying
to flush data in order to satisfy the fsync, and the VM already doing
writeback on it (though I can't see why writeback mode would be so much
better in this case.)

--Stephen





[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]