CPU usege too HIGH

[hugle]

JA> What is behind that server? If there are many servers/clients behind that
JA> server that require net access and that box functions as the gateway, then it
JA> is not really a surprisement that the loads are so high.

nothing behind the server, just clients... I have about 500 users, and
this BOX is doint NAT for all of them.

JA> Also, how many users are connected to that server? A few users won't bring
JA> that much load, but having twenty or thirty of them connected will bring
JA> quite a load to the CPU. It would make sense that Squid and NAT are using so
JA> much CPU time then.

JA> Your script may be small, but if again used by the same thirty people then
JA> perl will put quite a load on the server.
But this never used to happen for over a mnoth... i had the same count
of users. I have 3 internet connections.
Even now, when it is 1 o'clock (AM) when everybody sleeps (almost)
there is:
  PID USERNAME PRI NICE  SIZE    RES STATE    TIME   WCPU    CPU COMMAND
46087 root      63   0  9924K  9636K RUN    677:27 87.74% 87.74% natd
63795 nobody     2   0   305M   305M poll   176:15  5.08%  5.08% squid
60589 root       2   0   692K   388K sbwait 316:23  2.05%  2.05% natd
perl# uptime
12:49AM  up 22 days,  3:19, 1 user, load averages: 2.24, 2.15, 1.86

in last week load avarage at night was about 0.40 ....

with best wishes
Jarek
JA> Cheers,

JA> Jorn

JA> On Wednesday 25 February 2004 16:27, hugle wrote:
>> In last past days, users confirmed of big lag in games..
>> I looged into the server, and saw big proccessor usage there:
>>
>> last pid: 90449;  load averages:  5.55,  9.11,  6.74           
>>                          up 21+19:48:35  17:18:34 68 processes: 5 running,
>> 63 sleeping
>> CPU states: 29.6% user,  0.0% nice, 34.6% system, 35.8% interrupt,  0.0%
>> idle Mem: 321M Active, 1313M Inact, 286M Wired, 85M Cache, 199M Buf, 3488K
>> Free Swap: 4079M Total, 216K Used, 4079M Free
>>
>>   PID USERNAME PRI NICE  SIZE    RES STATE    TIME   WCPU    CPU COMMAND
>> 46087 root      62   0  8912K  8624K RUN    393:39 28.86% 28.86% natd
>> 63795 nobody    62   0   191M   190M RUN     22:38 22.61% 22.61% squid
>> 87001 root      -6   0  9288K  8756K piperd   0:04  7.57%  7.57% perl
>> 60589 root       2   0   672K   368K sbwait 297:12  0.24%  0.24% natd
>> 65212 root      66  19 17416K 15684K RUN     89:50  0.00%  0.00% mlnet-real
>>
>> perl# uptime
>>  5:24PM  up 21 days, 19:54, 2 users, load averages: 8.05, 11.78, 8.94
>>
>> i have a little sciprt running avery 5 minutes.. it uses perl, but..
>>
>> why natd is using so much resourses ?
>> there I have to dig the problem?
>> and why active memory is so low ?;/