On Wed, Feb 19, 2003 at 01:55:01PM -0000, Mark Cave-Ayland wrote:
> > How long did it take to get that trace? Also, what are file
> descriptors
> > 58,
> > 97 and 114?
>
> The trace lasted about a couple of minutes. I've listed the filesystems
> from /proc and uploaded them to
> http://www.infomapper.com/strace/fd.log.txt for you to look - I have a
> feeling you would find it useful to see the total number of files open
> in terms of their size and quantity....
58 = 55792/58153.15
97 = 55792/58153.54
114 = pg_xlog/000000A50000000E
So it's writing to WAL. I don't know if I'm sure about this but if it's
currently at file no 54 and you need to end up with 45 (each file 1GB) and
you started on file 69 that means you're over halfway. But I'm really not
sure.
> Agreed! If it ruins any caching then in my view it's something that has
> to change in order to keep performance. While there may be a penalty to
> pay on smaller tables, the benefits of caching would more than make up
> for the cost of going forwards - imagine how slow CPUs would be if
> everything was a cache miss....
I'd like to get one of the developers views on this.
--
Martijn van Oosterhout <kleptog@svana.org> http://svana.org/kleptog/
> Support bacteria! They're the only culture some people have.