Backend servers are correctly checked but request is made to local node only

From: Stefan Reger <sre#itscope.de>
Date: Mon, 21 Mar 2011 11:18:33 +0100


Hello,   I use the latest snapshot from 21/03 and there seems to be a bug with the following configuration.   Haproxy is running on 213.93.23.151:80 = 10.0.1.19:80. The availability check is correctly made to both of the backend nodes but the request itself is forwarded only to the local node (10.0.1.19:81). Hatop correctly shows the balancing to both nodes but in fact only one node receives and logs all of the requests.   To prove this I shut down the local node (10.0.1.19:81) and hatop shows a connection error to 10.0.1.18:81 since it thinks the node is up (availability check passed!) but tries to forward the request to 10.0.1.19:81 (which is down).   Could be a NAT problem!? Thank you for your help.   In 1.4 all works perfectly!     global                   #chroot /usr/share/haproxy                 user haproxy                 group haproxy                 daemon                 nbproc 1                 #debug                 #quiet   defaults                 log          global                 mode http                   timeout connect      5000                 timeout client      50000   timeout server      50000                   frontend public   bind *:80   default_backend nodes     backend nodes     option httpchk GET /rest/testAvailability   balance roundrobin     server js1 10.0.1.18:81 check inter 4000   server js6 10.0.1.19:81 check inter 4000       regards, Stefan Reger         Received on 2011/03/21 11:18

This archive was generated by hypermail 2.2.0 : 2011/03/21 11:30 CET