On Sat, May 16, 2009 at 09:44:53AM -0400, Tom Potwin wrote: > Hi Willy > > I checked the cfg files for both HAProxy and heartbeat, and they're the same > where they are supposed to be. You we're right about the SYSLOGD="-r" > setting. I didn't know I had to do that. I've attached a new copy of the > haproxy.log file I started when I restarted both LB1 and LB2 servers. It > shows from the time it works at 08:54, till it stops at 09:14. The log is > from LB1, and the same log on LB2 shows "May 16 09:14:58 lb2 -- MARK --" > when I couldn't get back into the stats. I still can get to my test local > web site even after the stats go away.
But there is no log after 9:11. Are you really sure that your browser is still going to the load-balancer's IP address ? Do you go there with its IP address or domain name ? Maybe you have multiple "A" records for the same host and your browser is rotating between them ? Could you also try connecting to LB1's own IP address instead of the shared IP address, so that at least we get a clue whether it's caused by heartbeat doing strange things or something else ? Right now I would say that some of your traffic does not even reach the load balancer :-/ You could even install tcpdump on LB1 and check for your incoming requests. Willy