3.2 Installation on Cluster nodes (FUJITSU Cloud Service K5 environment)

This section explains cluster node installation using the CLI installer in a FUJITSU Cloud Service K5 environment.

Note

Before installing PRIMECLUSTER, take a snapshot of the system disk.

Kernel header
Before installing this software, it is necessary to install the kernel header that supports OS of the system. Check if the kernel header is installed on the system by executing the following command:
```
# rpm -qi kernel-devel <Return>
```
If this command returns an error, or if a kernel source different from the version of the system OS is installed, use the following command to install the kernel source.
```
# yum install kernel-devel <Return>
```
Note
When using the yum command, refer to "2.1.2.6 Creating the Security Group for the Virtual Server Access" in "PRIMECLUSTER Installation and Administration Guide FUJITSU Cloud Service K5", or "2.1.2.7 Creating the Firewall Rule" in "PRIMECLUSTER Installation and Administration Guide FUJITSU Cloud Service K5" and check that the security groups and firewall rules have been set properly.
Installation of necessary OS packages
Use the following commands to check if the necessary OS packages are installed on the system.
```
# rpm -qi PyQt4-devel.x86_64 <Return>
# rpm -qi compat-libstdc++-33.x86_64 <Return>
# rpm -qi ebtables.x86_64 <Return>
# rpm -qi libXp.x86_64 <Return>
# rpm -qi openmotif.x86_64 <Return>
# rpm -qi openmotif22.x86_64 <Return>
# rpm -qi pam-devel.x86_64 <Return>
# rpm -qi rsh.x86_64 <Return>
# rpm -qi subversion.x86_64 <Return>
# rpm -qi openssl098e.x86_64 <Return>
```
If these commands return an error, use the yum command to install the package that returned an error.
Example) If PyQt4-devel.x86_64 returned an error.
```
# yum install PyQt4-devel.x86_64 <Return>
```
Note
- Before installing packages it is necessary to do the settings for Red Hat Update Infrastructure. For details, refer to "FUJITSU Cloud Service K5 IaaS Features Handbook."
- The following error message might be output and yum installation fails.
```
Protected multilib versions: ******.x86_64 != ******.i686
```
  Follow the procedure below and install the x86_64 package after first updating the i686 package.
  Example) If the libXp.x86_64 installation fails
```
# yum update libXp.i686 <Return>
# yum install libXp.x86_64 <Return>
```
  Specify the same version as the i686 package and install the x86_64 package.
Correcting the /etc/hosts file
To install the Web-Based Admin View, it is necessary to edit the IP address of the public LAN (and administrative LAN) as well as its host name (node name output in uname -n), as well as the host name (node name output in uname -n) for "127.0.0.1" (for IPv4), "::1"(for IPv6), used by the Web-Based Admin View in the /etc/hosts file.
1. Login to the system and become the root user.
```
$ sudo su - <Return>
Password:password
```
2. Delete the host name allocated to "127.0.0.1"(for IPv4) and "::1"(for IPv6) using vi(1) and allocate it to the IP address of the admin LAN.
  Before change)
```
# cat /etc/hosts <Return>
127.0.0.1 host-name localhost localhost.localdomain localhost4 localhost4.localdomain4
::1       host-name localhost localhost.localdomain localhost6 localhost6.localdomain6
```
  After change)
```
# cat /etc/hosts <Return>
127.0.0.1 localhost localhost.localdomain localhost4 localhost4.localdomain4
::1       localhost localhost.localdomain localhost6 localhost6.localdomain6
IP-Address host-name
```
Check the auto startup of the NetworkManager service.
PRIMECLUSTER does not support the NetworkManager service.
Check that automatic startup of the NetworkManager service is disabled.
Check that the setting of auto startup of the NetworkManager service is "off" using the following command.
```
# /sbin/chkconfig --list NetworkManager <Return>
NetworkManager	0:off	1:off	2:off	3:off	4:off	5:off	6:off
```
If there are runlevels that are "on", execute the following commands to stop the services.
```
# /sbin/service NetworkManager stop <Return>
# /sbin/chkconfig NetworkManager off <Return>
```
Copy the data from the DVD and deploy it in the environment where the installation is done.
Prepare separate environments where the DVD can be set and mounted and mount the DVD.
```
Copy source system # mount -t iso9660 -r /dev/<Device file name> <DVD-ROM mountpoint> <Return>
```
Copy the entire directory under <DVD-ROM mountpoint> to all the target systems copied to. Set the directory copied to as <Directory copied to>.
Note
When you copy to the target system, use the tar command to archive and make sure that the symbolic link is not made into an actual file.

Execute the CLI installer.

Prepare separate environments where the DVD can be set and mounted and mount the DVD.

# cd <Directory copied to>/Tool <Return>
# ./cluster_install -e CB-PG <Return>
Installation of PRIMECLUSTER started.
PRODUCT : PCLsnap
Installing package <FJSVpclsnap> ... finished.
       .
       .
The installation finished successfully.

When using GLS, execute the CLI installer again to install PRIMECLUSTER GL.

# ./cluster_install -e PCL-GLS <Return>
Installation of PRIMECLUSTER started.
PRODUCT : GLS
       .
       .
The installation finished successfully.

Download the latest PRIMECLUSTER patch(es) and update information file from Updatesite.
Apply the corrections to PRIMECLUSTER.
For how to apply changes, refer to "Software Maintenance" in "PRIMECLUSTER Installation and Administration Guide FUJITSU Cloud Service K5", and see the points of concern in the correction information file of each correction.