Re: performance testing and 503 errors

HAProxy is probably marking your servers as down when they fail the HTTP protocol checks.

In your config you specified option httpchk. According to the documentation, by specifying that, each server will receive an OPTIONS request on the URL "/" and only 2xx or 3xx HTTP response codes will be considered successful. I'm sure that's where the breakdown is. Is it possible that your Apache servers are denying OPTIONS requests, or something similar?

Also, try and comment out the option httpchk line and see if HAProxy stops marking your servers as down.

On 8/29/08, Martin Pegman <mpegman#mtld.mobi> wrote:
>
> Hi,
> I am having what I'm sure is a configuration issue with my first
> installation of haproxy and I would appreciate any pointers you may be able
> to give me.
>
> I originally had 2 apache2 servers running a php application behind a basic
> proxy that needs to access other internet sites for some content so the
> sessions are quite long. I have used JMeter to get a throughput of 3
> responses a sec and can handle 100+ concurrent connections.
>
> I have replaced this with a haproxy server in front of 5 apache servers
> (all different hosts) using the config below.
>
> However now when I run the same JMeter tests I almost immediately receive
> error 503 there is no server to handle your request errors.!
> And by immediately I mean before 100 urls have been tested.
>
> The apache servers never even appear to show any load so I cannot see why
> ha proxy would assume they cannot handle more connections.
>
> I'm running on Ubuntu 8.04 kernel 2.6.21.7-2 with HA-Proxy version 1.3.12
> 2007/06/17.
>
> I have read the available documentation and I cannot see how to resolve the
> issue. Any pointers or assistance would be greatly appreciated.
>
> Regards
> Martin P
>
> global #log 127.0.0.1 local0
>> log 127.0.0.1 local0 debug
>> log 127.0.0.1 local1 debug
>> #log localhost local0 debug
>> maxconn 50096
>> #debug
>> #quiet
>> user haproxy
>> group haproxy
>> daemon
>> nbproc 1
>>
>> defaults
>> mode http
>> clitimeout 150000 # maximum inactivity time on the
>> client side
>> srvtimeout 500000 # maximum inactivity time on the
>> server side
>> contimeout 500000 # maximum time to wait for a
>> connection attempt to a server to succeed
>> option httpclose # disable keepalive (HAProxy does
>> not yet support the HTTP keep-alive mode)
>> option abortonclose # enable early dropping of aborted
>> requests from pending queue
>> option httpchk # enable HTTP protocol to check on
>> servers health
>> option forwardfor # enable insert of X-Forwarded-For
>> headers
>> option clitcpka
>>
>> #stats
>> stats enable # enable web-stats at /haproxy?stats
>> stats auth admin:pass # force HTTP Auth to view stats
>>
>> ##Instant mobiliser
>> listen im_a_proxy 0.0.0.0:80 <http://0.0.0.0/>
>> balance roundrobin
>> option httplog
>> option dontlognull
>> option abortonclose
>> log 127.0.0.1 local0 debug
>> server im1 ip-10-250-9-238.ec2.internal weight 1 check
>> server im2 ip-10-250-9-219.ec2.internal weight 1 check
>> server im3 ip-10-251-69-220.ec2.internal weight 1 check
>>
> server im4 ip-10-250-8-44.ec2.internal weight 1 check
> server im5 ip-10-251-69-188.ec2.internal weight 1 check
>
>
>
Received on 2008/08/29 15:40