Hi, On Tue, Feb 03, 2009 at 08:40:45AM +0100, [email protected] wrote: > Morning, > > i have done some tests with haproxy and i need some precision regarding the > comportment above : > > am using: > - kernel 2.6.28.1 from kernel.org which include "TPROXY patch" > - haproxy -V > HA-Proxy version 1.3.15.3 2008/09/02 downloaded from binary repository on > the project website. > > i try to use configuration like this : > > listen TCP_XXXX 192.168.24.55:XXXX > mode tcp > balance roundrobin > retries 3 > contimeout 300000 > clitimeout 300000 > srvtimeout 300000 > > source 192.168.24.56 usesrc clientip > > server srva 192.168.24.12:XXXX check > server srvb 192.168.24.13:XXXX check > and this configuration working perfectly, the health test on servers is done > using 192.168.24.56 IP source and the "data flow" is sent using client IP : > at this point all is OK > > reading the documentation, it's seem possible to use those parameters like > the configuration below : > > > listen TCP_XXXX 192.168.24.55:XXXX > mode tcp > balance roundrobin > retries 3 > contimeout 300000 > clitimeout 300000 > srvtimeout 300000 > > source 192.168.24.56 usesrc 192.168.24.55 > > server srva 192.168.24.12:XXXX check > server srvb 192.168.24.13:XXXX check > > am waiting for the health test on servers done by 192.168.24.56 and "data > flow" by 192.168.24.55 ... > > or all connections to backend servers are done with 192.168.24.55 IP adresse > .... > > is it a bug, misconfiguration or misundertanting ?
At first glance, it looks like a bug to me, as the health-checks should use 192.168.24.56. I'll have a look at that. Cheers, Willy
