 | |||||||||||
| |||||||||||
 |
|||||||||||
| |
| ||||||||||||||||
| ShaoLin HA Cluster 1.0 Installation Manual and Operations Guide | ||
|---|---|---|
| Prev | Chapter 10. Cluster Administration | Next |
Changing hardware configurations requires offline operation, you may want to do this in your planned downtime in proper arrangement. unless the hardware is capable of hot-plug. The best way to do this is to add hardware peripherals to your standby server while the server is offline. This is the steps for adding or changing hardware configuration with minimal downtime. Generally, these steps works with most situation except for hard drives and NICs. Please refer to
Check your hardware peripheral is certified with your Linux distributor. Make sure it is supported by the distribution. If is good to prepare two piece of the same hardware to install at both servers to synchronize the hardware configurations for both clusters.
To install the new hardware, power-off the standby server. Install and configure your hardware peripheral properly.
Power on the standby server. At the boot loader (e.g. lilo or grub), choose to enter a test mode. At the console of Cluster Manager, check the status of the heartbeat and see if there is any problems.
If there is no problem, power off the standby server. At the active server, turn off the hardware detection service by using chkconfig kudzu off (if you are using a Redhat).
Power on your standby server to normal operation, at the console of Cluster Manager, manually take over the system.
After system failover, check the status of your applications and system availability. If everything are ok, you should run the hardware detection services manually by the command /etc/init.d/kuzdu start (if you are using Redhat). The hardware detection service should able to discover your newly installed hardware. After the hardware configuration is done, you should do appropriate configurations for your newly installed hardware. If your system failed to operate correctly after the failover, you should immediately roll back by take over the system using the previous active server.
After the operating system is configured and while the newly configured machine is on-line. You should take the original active server offline, and repeat the hardware installation (if necessary). After you have installed the hardware configurations, you should repeat step-3 but without the need of re-run any hardware detection service since it has been done by the other server already.
The problem of changing network configurations for a clustered system is that the clusters themselves uses network to perform inter-node communications. It is OK to change IP address of the clusters without the need of reconfiguration because the HA Cluster use non-IP Ethernet heartbeats. If you hare using IP address monitor, likely you will have to re-configure your clusters by re-run the Section 5.1
Another possibility of failure is adding an additional NIC to your system of the same type. If your new NIC adding to the system is same as the NIC's that is using in your system. Your existing driver which has configured for your system will also recognize the new NIC and automatically make it available for the system. You have a chance of re-ordering your Ethernet adapters identity in your system (e.g. eth0, eth1, eth2 etc.). If this happens, the kernel heartbeat daemon may use an incorrect Ethernet interface to heartbeat after the installation of the new NIC. To fix this, you will have to power off the standby server and re-configure heartbeat, see Section 5.3 for more information.
Generally, adding storage devices to HA Cluster will have not much difference with a standalone Linux server. You just have to make sure the newly added hardware will not cause the existing device name space to change (e.g. /dev/sda, /dev/sdb). If the reference has change, it is likely your system will failed to boot properly. It this happens, you should power off the standby server, manually resolve the problem and re-configure HA Cluster using the Section 5.1.
