history write time oddity

Alex Kiernan alexk at demon.net
Tue May 14 08:51:39 UTC 2002


Ben Rosengart <br+inn at panix.com> writes:

> I'm building some new news servers for Panix, based on INN 2.3.2 and
> what will be 2.3.3, and I'm seeing something that weirds me out.  On
> both the new feeder and the new readers, history write time as listed
> in the nightly reports is about a factor of 10 higher than it was on
> the old feeder and reader.
> 
> The old servers are both running "INN 2.3experimental 20-Nov-1998".
> They're built to use tagged-hash, could that be the difference?  Or
> has there been a change in the way history is written (I was thinking
> perhaps it used to be cached and now it isn't)?
> 
> I'm especially puzzled because the newer machines have much better
> I/O bandwidth than the old ones.  I'm pretty sure it's not a physical
> problem, as we've tried single disks (IDE and fibre channel) and
> striped and it doesn't seem to make a difference.
> 

Same OS on both boxes?

-- 
Alex Kiernan, Principal Engineer, Development, Thus PLC


More information about the inn-workers mailing list