| |
| |
| |
|
Page: 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
Comments:
<0> :D
<0> hey..vmstat is working :D
<1> vmstat -s
<1> what are the iowait cpu ticks?
<0> 0 IO-wait cpu ticks
<0> 717251 system cpu ticks
<0> i get from that there is no file activity
<0> ?
<0> ..and i have noticed this happens when there are a LOT of connections
<0> getting throgh the box
<1> hmmm
<0> now i'm talking over 6k connections both ways
<0> ..i thing torrents :D
<0> *think
<1> what network cards have you got?
<0> ..standards
<0> rtl8139too
<1> hmm there are better cards
<0> i know
<0> but i think NICs are not the problem
<1> thye also have a tendnacy to fall over under high loads
<0> firs i thoght there is a CPU problem
<1> e1000s are way better
<2> Does that driver support NAPI?
<1> even e100
<0> yeah it happened... but this NICs where more expensive than the first..guy at store saz those are better
<2> Heh.
<0> gcbirzan NAPI ?
<2> Yes, NAPI.
<1> 8139s are cheap and cheerful
<1> not for servers
<0> ...i dont know what napi is
<2> LD: What is your problem?
<3> Hmmmmm I'm trying to use powersave -U and it tells me that I need to have resume specified in my bootloader somehow, any ideas?
<0> well.. cpu load way to much
<1> what nics were in the cerelon?
<0> same
<1> k
<0> i think older 8139C chipset
<0> those are 8139D
<1> k
<1> c was common
<0> NICs are not the problem
<0> yeah I know...never had problems with those
<2> LD: How much traffic?
<1> probably not but they take more cpu than the e100s
<0> gcbirzan: Total 2.78MiB 4.38K 3.35MiB 4.45K
<4> celeron are bandwith eaters..lols
<0> now
<0> like 9k connections
<2> LD: Hm. What CPU?
<0> when io have many connections my box is dieing :(
<0> sempron 2800+
<1> hmm
<4> ld: what distro?
<0> gcbirzan i had FC3 also installed...same problems
<1> maybe the ports needs opening up
<0> kuto slack 10.1
<0> thing.. NO ports are blocked in FORWARD chain
<1> it could be some kernel tuning options
<0> only INPUT
<4> LD: you use nat in your networks?
<0> yes
<0> also
<0> NAT
<2> LD: Well, we have machines handling way more traffic than that, and doing just ifne.
<2> LD: But, then again, we use e1000.
<4> yeah me too
<0> :)
<2> Oh.
<2> NAT == bad.
<1> e1000 are good
<2> Connection tracking is a b*tch.
<0> gcbirzan i have over 5k rules in my firewall
<0> ..i know .. i had to increase conntrack
<4> e1000 is a gigabit nic?
<1> yes
<4> gigabit nick is slow if connected to 10/100 mbps hub
<2> Heh.
<1> who uses hubs these days?
<2> Nobody uses hubs.
<0> :) i do.. i also have old segments on BNCs
<4> eeer..switch
<2> kuto: That's... wrong.
<1> kuto i havnt noticed that
<0> gcbirzan I have over 2km of BNC network !
<0> :D
<4> thing: well, i encountered it already
<2> One problem I've noticed was with a Cisco 7600...
<4> especially using windows servers
<2> Connecting some kind of NICs (couldn't find out exactly what kind) directly into it leads to latencies.
<0> btw..while we are talking.. my box is down : Cpu(s): 0.0% user, 100.0% system, 0.0% nice, 0.0% idle
<0> "down"
<4> seems like gigabit nic's were not that compatible with 10/100 mb switches
<2> But, it had nothing to do with 100/1000.
<1> kuto often you have to lock the speed
<4> ld: have you tried changing nics?
<0> well.. no actually
<4> thing: using what?
<0> i really dont think is about nics
<0> 8139D
<1> kuto ethtool, miitool option in modules.conf
<1> auto-neg ****s
<1> LD probably not
<2> LD: dmesg | grep -i napi
<4> why not let the box optimized its speed, instead of using 3rd party?
<2> gcbirzan@QR-100:~$ dmesg | grep -i napi
<2> Intel(R) PRO/1000 Network Driver - version 5.6.10.1-k2-NAPI
<2> You want NAPI.
<0> heh no output
<2> NAPI is good.
<0> gcbirzan now u made me curios..what's NAPI ?
<2> Google.
<0> :)
<4> gyaah...
<0> heh..you are like my teacher.. he always sad: look it up !
<4> lets kill google, so that everything is manual eh!
<1> ladybyte, napi
<1> ladybyte, google for napi
<5> google can't find napi
<1> :/
<0> thing lady has the module off..i think
<4> ladybyte, bark!
<0> gcbirzan: i've killed all my trafic and my box is breathing a little.. questin: can a flood put my box in 100% cpu ?
<4> ld: you have webserver running on that box?
<0> not yet
<0> 13 houres uptime..of the OS :)
<4> what does your box running, ICS?
<0> yes..main goal
<6> hive us some vmstat info
<0> also DC hub, squid, CS serevr, web, main, bind...
<4> is it already running?
<0> also DC hub, squid, CS serevr, web, main, bind...none working *yet*
<6> hive us some vmstat info LD
<0> cant p*** firewall putting 100% CPU issue
<0> r b swpd free buff cache si so bi bo in cs us sy id wa
<0> 0 0 0 558460 117460 80380 0 0 5 5 712 38 1 24 75 0
<6> faults in 712?
<2> LD: Yes.
<0> faults ??
<6> hit vmstat 1 and give us a better view of it
<6> thats interr./s
<0> a minute ago i had ip_forward stop them :).. let my box live a little.. now its running again and i see connections go wild again
<0> sergiu25 0 0 0 558456 117460 80380 0 0 0 76 5128 44 0 59 41 0
<6> what is 5128?
<0> in
<6> that's the problem
<0> system: in
<0> i know..look:
<6> you have a lot of traffic probably
Return to
#linuxhelp
or
Go to some related
logs:
pda sync whit ubuntu
orkut scripts
jsf passing parm to page
#networking
nannu
#london
#linux
#c
#teens
#apache
|
|