On Thu, 06 Sep 2001 16:31:27 +0200, Sheldon Hearn wrote:
> Obviously, I wouldn't _dream_ of enabling no_fsync in queue runners,
> only in queue injectors (exim -bS).
Well, I got curious and gave no_fsync a whirl with 200 queue runners.
They chewed through my 160,000 message queue at a rate of between 50Kmph
and 75Kmph [1], compared against the usual 18Kmph to 25Kmph that I see
without no_fsync.
The only thing that slowed me down were the incoming bounces, which are
handled by instances of exim that don't use no_fsync.
Given my fairly unusual configuration, where the entire spool is going
to be blown away anyway on failure, this is quite tempting. :-)
