Hi John,

On Wed, Mar 30, 2011 at 12:42:28PM +0100, John Helliwell wrote:
> Hi all,
> 
> We are using HAproxy as a L7 load balancer on our busy phpBB3 based forums,
> and we are finding it excellent.
> We would like to upgrade to haproxy-1.5-dev in order to use the rate
> limiting as described in http://blog.serverfault.com/post/1016491873/
> 
> However, this has not been possible, due to the fact that in haproxy-1.5-dev
> the httpchk fails, and all backend servers are marked down.
> 
> Mar 30 11:30:19 localhost haproxy[24995]: Server to-slow-queue/forum-admin1
> is DOWN, reason: Socket error, info: "Invalid argument", check duration:
> 0ms. 1 active and 0 backup servers left. 0 sessions active, 0 requeued, 0
> remaining in queue.
> 
> Also, it cannot syslog due to
> 
> [ALERT] 088/120607 (26987) : sendto logger #0 failed: Invalid argument
> (errno=22)
> 
> The relevant config extract for this is
> defaults
>    option httpchk GET /health.html
> 
> backend webservers
>     server forum-web1 172.16.50.192:80 minconn 6 maxconn 10 weight 12 check
> inter 2s fastinter 500ms downinter 10s slowstart 60000
> 
> The machine is a Sun Fire V125 with an UltraSPARC-IIIi processor, compiled
> locally using gcc version 3.4.3 using
> gmake TARGET=solaris ARCH=ultrasparc USE_PCRE=1
> 
> I can provide full haproxy.cfg if required.
> 
> Is this a known issue? I hope someone can assist.

It's still unknown. Are you seeing that only with 1.5-dev ? Also the fact
that it fails on the syslog too is strange. It looks like it doesn't like
one of the flags passed to send().

Could you run it through truss and send the output ?

Thanks,
Willy


Reply via email to