Installation d'un cluster Red Hat 4
To install Red Hat Enterprise with Cluster Suite, take CDs or DVD and boot on it.
We are going to see step by step the interesting things.
- Language
Put the English language and not French to have more verbose during futur problems.
- Keyboard
Keep the French keyboard
- Partitioning (do the same partitioning on all the cluster nodes)
| Location | Type | Size | LVM Name |
|---|---|---|---|
| /boot | ext3 | 128 Mo | NO LVM !!! |
| LVM | Group | Full Size (space left) | VolGroup00 |
| / | ext3 | 4096 Mo | racine |
| /var | ext3 | 4096 Mo | var |
| /usr | ext3 | 10240 Mo | usr |
| /tmp | ext3 | 2048 Mo | tmp |
| swap | 4096 Mo | swap | |
| /home | ext3 | Space left | home |
- List of packages installation
base base-x kde-desktop or gnome-desktop editors (only vim) mail-server (- spamassassin) development-tools compat-arch-development admin-tools (+ sysstem-boot-config) system-tools compat-arch-support
- After rebooting install those (Warning do attention for SMP, if not SMP servers, do not install it install other relative pacakges) packages in the cluster suite (cdrom/RedHat/RPMS)
ccs cman cman-kernel-smp dlm dlm-kernel-smp fence gulm ipvsadm magma magma-plugins perl-Net-Telnet rgmanager system-config-cluster
or do that line directly
rpm -ivh --aid ccs-1.0.7-0.x86_64.rpm cman-1.0.11-0.x86_64.rpm cman-kernel-smp-2.6.9-45.2.x86_64.rpm dlm-kernel-smp-2.6.9-42.10.x86_64.rpm fence-1.32.25-1.x86_64.rpm gulm-1.0.7-0.x86_64.rpm ipvsadm-1.24-6.x86_64.rpm magma-1.0.6-0.x86_64.rpm magma-plugins-1.0.9-0.x86_64.rpm perl-Net-Telnet-3.03-3.noarch.rpm rgmanager-1.9.53-0.x86_64.rpm system-config-cluster-1.0.27-1.0.noarch.rpm dlm-1.0.1-1.x86_64.rpm
A little update might not be so bad (facultative)
up2date -u
Modify the /etc/hosts like that on each nodes :
127.0.0.1 localhost.localdomain localhost 192.168.0.242 secondary.cluster.net secondary 192.168.0.241 primary.cluster.net primary
Here is an example of the cluster.conf file in /etc/cluster
<?xml version="1.0"?> <cluster config_version="13" name="alpha_cluster"> <fence_daemon post_fail_delay="0" post_join_delay="3"/> <clusternodes> <clusternode name="primary.cluster.net" votes="1"> <fence> <method name="1"> <device name="fence1" nodename="primary.cluster.net"/> </method> </fence> </clusternode> <clusternode name="secondary.cluster.net" votes="1"> <fence> <method name="1"> <device name="fence2" nodename="secondary.cluster.net"/> </method> </fence> </clusternode> </clusternodes> <cman expected_votes="1" two_node="1"/> <fencedevices> <fencedevice agent="fence_manual" name="fence1"/> <fencedevice agent="fence_manual" name="fence2"/> </fencedevices> <rm> <failoverdomains> <failoverdomain name="myfailoverdomain" ordered="0" restricted="1"> <failoverdomainnode name="primary.cluster.net" priority="1"/> <failoverdomainnode name="secondary.cluster.net" priority="1"/> </failoverdomain> </failoverdomains> <resources> <ip address="192.168.0.238" monitor_link="1"/> <fs device="/dev/sdb1" force_fsck="0" force_unmount="0" fsid="44028" fstype="ext3" mountpoint="/mnt/cluster" name="disk2" options="" self_fence="0"/> </resources> <service autostart="1" domain="cluster" name="clusterdev"> <ip ref="192.168.0.238"/> <fs ref="disk2"/> <script file="/etc/init.d/ntpd" name="ntp"/> </service> </rm> </cluster>
And of course don't forget to run services
service ccsd restart service cman restart service fenced restart service rgmanager restart
If you want to stop cluster, do it in the reverse order
service rgmanager stop service fenced stop service cman stop service ccsd stop
It's very important to stop in this order !
