On Sun, 23 Jan 2011 04:20:04 +0700, Insan Praja SW <insan.pr...@gmail.com>
wrote:
Hi,
On Sun, 23 Jan 2011 01:54:32 +0700, Willem Dijkstra <w...@xs4all.nl>
wrote:
On 01/22/2011 01:27 AM, Insan Praja SW wrote:
Seems obvious that symux isn't detecting rollover properly for
whatever
variable you are seeing a graph "spike". It should be fairly easy for
them to fix if you report it. The fact that it affects 64bit and not
32bit counters is a damn good clue.
symux reports measurements to rrdtool, and rrdtool tries to detect
rollovers. Without looking at any source I would guess that the new
pfctl -f zeros the queue stat counts, which would lead to rrdtool
making a false overflow detection. This should affect both 32/64 bit
archs.
^^^^^^^^^^^^
In my last email, all 4.8 platform makes them same behavior.
Also note that this would be hard to fix in symon/symux; pfctl is right
to reset the stats, and rrdtools overflow handling is nice for when you
Comparing to systat(1) at queue view, when "pfctl -f /etc/pf.conf" was
executed, the counters went "*", is it zero-ed?
have overflows. I don't see a simple way of second guessing either of
them in the symon/symux source.
So the solution (right now) would be using rrdtool overflow handling.
We already contacted the maintainer, still no reply though.
^^^
This is the first mail I received about this subject, or are were you
talking about someone else?
Sorry Will, I'am talking about someone else, with less english writing
skill :D
Cheers,
Willem
Thanks,
Insan Praja
Thanks,
Insan Praja
--
Using Opera's revolutionary email client: http://www.opera.com/mail/
