Re: Virtual server not serving?

Hi Brendan,


> I'm testing with just one real server and have stopped the iptables
> service on both servers for testing purposes. 

in my testing environment, I only use the nat-Rule, no packet filtering
rules. Later in production (if I ever get it to this state) the whole
system will behind a firewall, so I don't mess with iptables-rules on
the routers. Again, naturally, except the nat-rule, because I'm going to
use lvs_nat. This is recommended, as long as I don't have access to the
production real-servers.


> The pulse service is starting and everything appears OK but the
> requests are not getting through to the real server, the http
> monitoring requests are though (can see them in access log):

That's what I am seeing, too. But from my networking knowledge, there
must be some kind of binding the virtual server outer ip to the service
that should be balanced. So, in "netstat -na |grep LISTEN" I expect
something like: 192.168.100.100:80 but there isn't. The only listening
ports are the sshd and the piranha-gui. :-(

So every time I try to connect to the virtual server with http, I get:
Connection refused. Routing should be ok, because I can ping the virtual
IP and all real IPs (routers and realservers) and a traceroute from
outside to the realserver-ips is no problem.


> The annoying thing is that I managed to get it working prior to
> shiping the servers to ISP, had to change ip's and gateway on eth0.
> Also fiddled with the httpd.conf though.

That's real bogus. Did you check the lvs.cf file with the new
networking-data?


> Are you starting the pulse service prior to bringing up the web
> servers?

I currently don't have access to the machines, they are at work. But I
think I start first lvs, then pulse, then the piranha-gui. I'll check
this tomorrow. The realservers are always running and I can connect to
each of them from each router with "telnet real-ip 80".

In addition, the nanny processes can reach the realservers and check for
load via uptime.


> I've noticed a debug_level in the lvs.cf file may be worth a
> look.

Somewhere (I think in the manpage) I found that this debug option is not
supported yet. :-(


> Check your httpd.conf file.

Hm, what settings are there that can interfere with the routers?


> Have you checked back in the archive for similar issues?

Yes. I've found some information but noone seems to have this problem.
Quit strange.


Today, I thought about it and identified some possible caveats:

I installed piranha and ipvsadmin via source-rpms. The time I compiled
piranha, there was an outdated ipvsadmin on the machine. Perhaps I
should recompile piranha with the new ipvsadmin?

If this won't work, I'm going to fiddle a bit with configuration, remove
the backup router from config, etc...

If this won't help either, I will configure the whole thing the lvs-way,
e.g. without piranha.

But it would be great if I get the ultimate hint before all this work.

And I wish you the solution of your problem, too.

 -volker