Dear Eliezer
as to my tests on 3.1 it can handle about 100Mbps with single core with
about of 40% cpu load on intel xeon or AMD X6 without the CPU spikes
i have 4 servers 3 on old gentoo kernel and one on newer ( portage,
kernel and libtool ) the only new one has the spikes
i had to add the GCC flags on the new one due to an error in compile
g++: warning: switch '-fhuge-objects' is no longer supported
tools.cc: In function 'void no_suid()':
tools.cc:778:14: error: ignoring return value of 'int setuid(__uid_t)',
declared with attribute warn_unused_result [-Werror=unused-result]
tools.cc: In function 'void enter_suid()':
tools.cc:756:14: error: ignoring return value of 'int setuid(__uid_t)',
declared with attribute warn_unused_result [-Werror=unused-result]
cc1plus: all warnings being treated as errors
make[3]: *** [tools.o] Error 1
make[3]: Leaving directory `/usr/squid/src'
make[2]: *** [all-recursive] Error 1
make[2]: Leaving directory `/usr/squid/src'
make[1]: *** [all] Error 2
make[1]: Leaving directory `/usr/squid/src'
make: *** [all-recursive] Error 1
adding cflags solved the problem ...
so the new one is the only one with the cflags
is there any effect for the cflags on performance and gcc errors (
while loops ) which causes the CPU to spike to 100% ????
best regards
On 2014-03-31 05:57, Eliezer Croitoru wrote:
On 03/31/2014 11:41 AM, a.afach@xxxxxxxxxxxxx wrote:
AMD Phenome 2 X6 / 16 GB ram / 1 TB SSD /2 TB SATA/
Why 3.1.19? have you considered that one cpu cannot take the load by
any chance?
i know that one core is working for squid but the spikes occurs not
only
in peak times and after spike the CPU return to 30% or 20%
i can see the error because of gcc or while loops
CC=x86_64-pc-linux-gnu-gcc CFLAGS=-O2 -pipe -m64 -mtune=generic
LDFLAGS=-Wl,-O1 -Wl,--as-needed CXXFLAGS= --cache-file=/dev/null
--srcdir=.
is this cflags true or ????
It's not the issue we are talking about.
There are couple aspects to the same issue.
While some expect some results to be achieved there is a limit to each
CPU\CORE.
It can handle about 50MBps of traffic and which is a maximum and
after that you must use more cores.
now squid.conf is relevant to understand how you use it.
Also what is the amount of traffic that the server takes?
Eliezer
