[Xapian-discuss] Omega datevalue search fails during scriptindex flush

[James Aylett]

On Fri, Dec 14, 2007 at 09:56:24AM +0100, Steije van Schelt wrote:

> I've tried running the query with ./omega, it takes about 68 seconds
> for the query to complete. The machine is dedicated to Xapian and
> has 2 gb of RAM. Average disk-usage is about 1.5mb/s, which is not
> too much isn't it? (scriptindex was running aswell)

Depends hugely on your disk(s). You want to look at iostat and find
out what the disk utilisation is, and get some idea of your cache
throughput and buffer usage. One way of doing this is to fling the
net-snmp and cacti packages onto the system and gather some background
data (for a couple of hours), then run the thing that causes you the
problems (you'll probably want to increase the polling interval in
cacti from 5 minutes so you can actually capture this - or script the
search to happen repeatedly for 15 minutes or so).

There are other approaches, but that's the easiest to describe (and I
don't know FreeBSD well, but I know those packages exist :-). Good luck!

J

-- 
/--------------------------------------------------------------------------\
  James Aylett                                                  xapian.org
  james at tartarus.org                               uncertaintydivision.org