Re: Low performance in solaris 9

From: <manuelsspace-listas#yahoo.com>
Date: Mon, 5 May 2008 08:56:07 -0700 (PDT)


Hello List,

   The -Vd output is:

root#sunexplor # ./haproxy -f digitel.cfg -Vd

[WARNING]
125/093451 (2187) : parsing [digitel.cfg:34]: keyword 'redispatch' is deprecated, please use 'option redispatch' instead. Available polling systems :

       poll : pref=200,  test result OK
     select : pref=150,  test result OK

Total: 2 (2 usable), will use poll.
Using poll() as the polling mechanism.

root#sunexplor # Message from syslogd#localhost at Mon May 5 09:34:54 2008 ... localhost haproxy[2188]: proxy backend3 has no server available!  

Message from syslogd#localhost at Mon May 5 09:34:55 2008 ... localhost haproxy[2188]: proxy backend2 has no server available!

There is no connection errors in syslog

The
test was stablish as 2 scenarios: a) test using haproxy, 1 client and 1 backend. b) the same client and the same backend but w/out haproxy. In this case, 200 threads x 200 seconds loading 10k html file each time We have observed 100 connection at top load + 100 queue in haproxy console in apache3. I don't know if this is associated to 200 poll available as -Vd display

Which client?, pylot 1.1, command line mode (http://www.pylot.org)

Haproxy statistics:

#pxname,svname,qcur,qmax,scur,smax,slim,stot,bin,bout,dreq,dresp,ereq,econ,eresp,wretr,wredis,status,weight,act,bck,chkfail,chkdown,lastchg,downtime,qlimit,pid,iid,sid,throttle,lbtot,
backend3,FRONTEND,,,0,0,2000,0,0,0,0,0,0,,,,,OPEN,,,,,,,,,1,5,0,,,
backend3,light1,0,0,0,0,,0,0,0,,0,,0,0,0,,DOWN,1,1,0,0,1,563,563,,1,5,1,,0,
backend3,light2,0,0,0,0,,0,0,0,,0,,0,0,0,,DOWN,1,1,0,0,1,562,562,,1,5,2,,0,
backend3,light3,0,0,0,0,,0,0,0,,0,,0,0,0,,DOWN,1,1,0,0,1,562,562,,1,5,3,,0,
backend3,light4,0,0,0,0,,0,0,0,,0,,0,0,0,,DOWN,1,1,0,0,1,562,562,,1,5,4,,0,
backend3,BACKEND,0,0,0,0,2000,0,0,0,0,0,,0,0,0,0,DOWN,0,0,0,,1,562,562,,1,5,0,,0,
backend2,FRONTEND,,,0,0,2000,0,0,0,0,0,0,,,,,OPEN,,,,,,,,,1,4,0,,,
backend2,tux5,0,0,0,0,,0,0,0,,0,,0,0,0,,DOWN,1,1,0,0,1,562,562,,1,4,1,,0,
backend2,tux6,0,0,0,0,,0,0,0,,0,,0,0,0,,DOWN,1,1,0,0,1,561,561,,1,4,2,,0,
backend2,tux7,0,0,0,0,,0,0,0,,0,,0,0,0,,DOWN,1,1,0,0,1,561,561,,1,4,3,,0,
backend2,BACKEND,0,0,0,0,2000,0,0,0,0,0,,0,0,0,0,DOWN,0,0,0,,1,561,561,,1,4,0,,0,
backend1,FRONTEND,,,0,200,2000,22789,4626167,240310005,0,0,0,,,,,OPEN,,,,,,,,,1,3,0,,,
backend1,apache3,0,0,0,100,100,22789,4626167,240310005,,0,,0,0,0,,UP,1,1,0,0,0,565,0,,1,3,1,,22789,
backend1,BACKEND,0,100,0,200,2000,22789,4626167,240310005,0,0,,0,0,0,0,UP,1,1,0,,0,565,0,,1,3,0,,22789,
switch,FRONTEND,,,0,205,2000,22789,4626167,240310005,0,0,0,,,,,OPEN,,,,,,,,,1,2,0,,,
switch,backend1,0,0,0,200,,22789,4626167,240310005,,0,,0,0,0,,no check,1,1,0,,,,,,1,2,1,,22789,
switch,backend2,0,0,0,0,,0,0,0,,0,,0,0,0,,no check,1,0,1,,,,,,1,2,2,,0,
switch,backend3,0,0,0,0,,0,0,0,,0,,0,0,0,,no check,1,0,1,,,,,,1,2,3,,0,
switch,BACKEND,0,0,0,200,2000,22789,4626167,240310005,0,0,,0,0,0,0,UP,1,1,2,,0,565,0,,1,2,0,,22789,
frontend,FRONTEND,,,1,227,2000,22795,3512241,240390316,0,0,0,,,,,OPEN,,,,,,,,,1,1,0,,,
frontend,switch,0,0,0,204,,22789,3509506,240310005,,0,,0,0,0,,no check,1,1,0,,,,,,1,1,1,,22789, frontend,BACKEND,0,0,0,204,2000,22789,3512241,240390316,0,0,,0,0,0,0,UP,1,1,0,,0,565,0,,1,1,0,,22789,

No errors at all. errors should be at SO level but the cpu usage is some times up to 100%

root#sunexplor
# ps -fl -p 2188

 F S      UID   PID  PPID  C PRI NI     ADDR     SZ    WCHAN    STIME TTY      TIME CMD
 8 S     root  2188     1  0  40 20        ?   2916        ? 09:34:52 ?        2:47 ./haproxy -f digitel.cfg -Vd

Backend statistics:
Red Hat Linux release 7.2 (Enigma)
Linux 2.4.9-13
Apache 1.3.37

At the moment that the client reports several connection errors the back end was as...

 10:00am up 153 days, 19:17, 10 users, load average: 2,26, 1,84, 1,45 273 processes: 265 sleeping, 6 running, 2 zombie, 0 stopped CPU states: 0,2% user, 0,1% system, 0,0% nice, 0,0% idle

Mem:   126240K av,  122424K used,    3816K free,     384K shrd,   20188K buff
Swap:  779112K av,   79932K used,  699180K free                   34560K cached

  PID USER     PRI  NI  SIZE  RSS SHARE STAT %CPU %MEM   TIME COMMAND
 9021 sg5       17   5  4272  692   616 R N  89,8  0,5  7132m Dialogic_DRV
30600 sg5       12   0  1184 1180   832 R     3,7  0,9   0:00 top
29647 sg5        9   0  3820 3712  2088 S     0,9  2,9   0:00 httpd
30272 sg5        9   0  2156 1544  1360 S     0,9  1,2   0:00 httpd
30320 sg5        9   0  2172 1560  1376 S     0,9  1,2   0:00 httpd
30363 sg5        9   0  2188 1576  1392 S     0,9  1,2   0:00 httpd
30382 sg5        9   0  2188 1576  1392 S     0,9  1,2   0:00 httpd
30386 sg5        9   0  2188 1576  1392 S     0,9  1,2   0:00 httpd


Haproxy config:

global         

daemon

        
maxconn 1000      # warning: this has to be 3 times the expected 
value!
        log 

localhost local0

defaults         

mode http         

balance roundrobin         

option dontlognull         

option httpclose         

retries 1         

redispatch

        
maxconn         2000
        
contimeout      5000
        
clitimeout      50000
        
srvtimeout      50000
        

stats enable  

listen
backend1 127.0.0.1:8001

        
log     global
        
option  httplog
        
capture request header X-Forwarded-For len 15
        
option  httpchk /index.html
        

server apache1 10.27.40.121:80 maxconn 100 check inter 2000 fall 3         

server apache2 10.21.17.206:80 maxconn 100 check inter 2000 fall 3         

server apache3 10.27.40.81:80 maxconn 100 check inter 2000 fall 3
... Others included in example which are related to unavailable listeners

Haproxy
will be used internally balancing web services. 300 users expected + up to 100 threads running in a weblogic requesting services or smpp (up to 100tps aprox)

Thanks,
Manuel Soto
Get Firefox!

On Fre 02.05.2008 09:23, manuelsspace-listas#yahoo.com wrote:
>
>> I'm a newbie in haproxy testing this product in Solaris 9.

> As fer as I can see in the compilation list no errors or warning was
> displayed. How can I verify if haproxy is ussing epoll o the best
> available opcion?

Sorry wrong platform!

epoll is only available on linux based system.

>$ make TARGET=solaris CPU=ultrasparc

[snipp]

Sorry but I can't see the output of

./haproxy -f examples/haproxy.cfg -Vd

With which tool do you make your tests?
What is your backend?
How looks your haproxy.conf?
Have you any errors in the haproxy logs?

BR

Aleks Received on 2008/05/05 17:56

This archive was generated by hypermail 2.2.0 : 2008/05/05 18:00 CEST