|
I have an UltraMonkey 3 streamlined load balancing system. There are only
two servers in this setup. The servers run as both directors and web/ftp
servers. The LVS manages about 115 IPs shared on both servers.
The servers used to work well and ran load balanced and failed-over
properly. Somewhere in tweaking the system recently however, I'm finding
that it no longer will allow both servers to run services at the same time.
The heartbeat is working ok, talking to each other and reporting no
problems, as is the ldirector only running on one server at a time. I can
have one machine running heartbeat and the other not however and the problem
still happens.
If I disable the interface (eth0) of one server, the problem goes away.
I had thought initially this was related to the problem of http and ftp on
the same server (which I've seen mentioned a couple of times on the mailing
list but can't find much of anything on) but when running on only one
director and webserver the http and ftp services coexist on the same server
(along with ldirectord and heartbeat) quite well. The IP with ftp does not
also run http on it.
When both servers are running heartbeat and have their public interfaces up
(and are freezing on site loads), an "ipvsadm -lcn" shows many connection
starts (SYN_RECV), but little to no established connections, idled out
connections, closed, or waiting connections.
/var/log/ha-log and /var/log/ha-debug show no apparent problems (but I'll
post them if it will help). Neither does ldirectord.log
So where do I start looking to see why the servers do not appear to be
responding to requests?
_______________
Dan Brown
danb@xxxxxx
|
