One of the installed processor boards was a dual Pentium III 550 MHz.
Selecting the SMP kernel towards the end of the installation process
allowed both processors to be detected on the following boot up.
Turbocluster Software
Our test environment is on a network using the Class B network
10.1.0.0/16. The three nodes in the cluster use the ip addresses
10.1.2.21, 10.1.2.22, and 10.1.2.23. The "virtual server" was assigned
the IP address of 10.1.2.25. These ip addresses all resolve back to the
hostnames labs021, labs022, labs023, and labs025, respectively.
Following installation, we modified the included
/etc/turbocluster.conf.new file on labs023 to reflect our setup. Here is
the output from diff comparing the original file and our modified file:
---------------------------------------------------------------
52,54c52,53
< # Service telnet tcp:23 xCheck
< # Service user udp:1024 yCheck
< # Service ftp tcp:22 none
---
> Service telnet tcp:23 none
> Service ftp tcp:21 none
63,65c62,64
< # Server server1 server1.domain.com tunnel
< # Server server2 server2.domain.com local
< # Server server3 192.168.1.3 direct noping
---
> Server server1 labs023.linuxcare.com direct
> Server server2 labs022.linuxcare.com tunnel
> Server server3 labs021.linuxcare.com tunnel
75,77c74,76
< # AddServer server1 80:http/1 telnet/2 ftp/2
< # AddServer server2 80:http/1 5000:user/1 ftp/1
< # AddServer server3 80:http/1
---
> AddServer server1 80:http/1 telnet/1 ftp/1
> AddServer server2 80:http/1 telnet/1 ftp/1
> AddServer server3 80:http/1 telnet/1 ftp/1
94,95c93,95
< # AddRouter router1.domain.com
< # AddRouter server1.domain.com
---
> AddRouter labs023.linuxcare.com
> AddRouter labs022.linuxcare.com
> AddRouter labs021.linuxcare.com
108c108,109
< VirtualHost virtual1.domain.com
---
> VirtualHost labs025.linuxcare.com
> MailTo ned@linuxcare.com
After saving our changes to the file /etc/turbocluster.conf, we ran the
program turbonetcfg and selected the menu option Turbocluster NFS
Config. Following the steps in this menu-based program, we configured
labs023 to be the NFS server for the cluster. We then ran the program
turbocluster_sync, which uses nfs services to copy the configuration
file to the other nodes in the cluster, and then starts or restarts the
turbocluster services on the other nodes.
Examining the log file /var/log/turboclusterd.log after the turbocluster
services start shows output such as:
1999-11-09 04:29:47 Starting turboclusterd version 4.0 for kernel 2.2.x
1999-11-09 04:29:47 Config file is /etc/turbocluster.conf
1999-11-09 04:29:47 Broadcast address is 10.1.2.255, network mask is 255.255.255.0
1999-11-09 04:29:47 This computer is server!
1999-11-09 04:29:47 Setting up module ipip and interface tunl0!
1999-11-09 04:29:47 Loading kernel module ipip
1999-11-09 04:29:47 Interface tunl0 127.255.255.0 up
1999-11-09 04:29:47 Interface tunl0:0 10.1.2.25 up
1999-11-09 04:29:47 Interface tunl0:0 10.1.2.25 up
At this point, we could verify that each of the three services being
provided by the cluster, telnet, ftp, and http, were working correctly.
We used the telnet program to test both ftp and telnet services. We
verified by entering a command such as 'telnet labs025 21' that the
connection request may actually be serviced by any one of the three
nodes in the cluster. This was determined by reading the name of the
actual server name in the connection banner displayed by the telnet
program. The same method was used to test telnet.
To verify that http connections to labs025 could be serviced by any node
in the cluster, we simply generated different html content on each node.
By examining the content displayed in the web browser, we could
determine which node was servicing the conne
