Well, changing syslog-ng to use the bulk loading API for various databases would probably increase performance a lot.
But that's a nontrivial code change, and would probably not work through libdbi.
On Wed, Oct 15, 2014 at 10:56 AM, Balazs Scheidler <bazsi77@gmail.com> wrote:
>
> It immediately runs the query, flush_lines() controls the transaction size,
> which can be enabled using
> flags(explicit-commits).
Ok, I see and I can confirm noticeable improvements on a system
under heavy I/O and CPU load using explicit-commits (or better: I see
it gets a lot worse disabling it, since I already had it on ;))
Plus, I'm seeing minor improvements with session_statements set
to "PRAGMA synchronous=OFF; PRAGMA count_changes=OFF;"
Beside this, has anyone found other ways to improve writing performances
using syslog-ng with sqlite?
On (physical) systems with high load, I've noticed that the system is much more
responsive if the elevator is set to noop or deadline instead of cfq.
But I guess this is highly dependent on the specific circumstances of
the system.
Thanks!
--
Davide Alberani <davide.alberani@gmail.com> [PGP KeyID: 0x465BFD47]
http://www.mimante.net/
______________________________________________________________________________
Member info: https://lists.balabit.hu/mailman/listinfo/syslog-ng
Documentation: http://www.balabit.com/support/documentation/?product=syslog-ng
FAQ: http://www.balabit.com/wiki/syslog-ng-faq