Last call for 2.4.3 problems

Russ Allbery rra at stanford.edu
Mon Mar 20 00:07:36 UTC 2006


Ulrich Schmid <schmid at switch.ch> writes:

> Sorry, something went wrong as I did the ctlinnd stathist test.  It's
> working, but my box needs with 2.4.3 more time for history lookup and
> history write as with 2.4.1. These results in about 10% less overall
> performance (incoming volume/articles) on my box.

I'm afraid I have no idea on this.  I reviewed the relevant changes as
near as I could figure them out, and I don't see anything that should make
history writes any slower.  The main change in this area is to avoid
fseeko and ftello to move around the history file since they do a flush
and instead use an internally maintained offset, but this should actually
speed things up and make fewer system calls.  The code was also modified
to actually *do* something for dbzsync, but history sync isn't a
substantial component of your timings either before or afterwards.

-- 
Russ Allbery (rra at stanford.edu)             <http://www.eyrie.org/~eagle/>

    Please send questions to the list rather than mailing me directly.
     <http://www.eyrie.org/~eagle/faqs/questions.html> explains why.


More information about the inn-workers mailing list