Contents
Failover HOWTO
This document describes how to change which machine in a redundant pair is actively serving requests. There are three different mechanisms for different clusters.
This document only describes the simple stuation where the active machine crashes and you want to engage the backup machine. This document does not attempt to describe the procedures for safely causing a failover when both machines are functional. In the case where both machines are functional, you want to be absolutely sure that the backup machine has a perfectly up-to-date copy of all user data before doing the failover.
![/!\](/~dmartin/wiki/modern/img/alert.png "/!\"){kind=small}
The backup machine automatically copies all user data from the active machine. If you switch to the backup machine and the formerly-active machine somehow comes back to life, all user changes since the crash will be lost when the now-active machine copies all the user data from the formerly-active machine.
When I say "Log on as root" I usually mean you should log on as your own user account and then use sudo to become root, like this "sudo su". Sudo asks for your password, not the root password. If sudo doesn't work, I have the root passwords in the safe in my office.
Firewalls
My firewalls are gone, so you shouldn't need to worry about anything firewally.
Arthur, Edison & Tintin
These systems do not do automatic failover. If you want the reserve machine to go live, you must do the failover manually.
These clusters use a single IP address for the official site. For instance, Arthur's official IP address is 168.156.85.16 and either arthur1 or arthur2 may use that IP address to function as the official server.
Arthur2, Edison2 & Tintin2 are the default running hosts. When they boot up, they start working on the live IP addresses. For Arthur1, Edison1 or Edison2 to become live, they have to steal that IP with gratuitous arp.
Switching from Arthur2/Edison2 to Arthur1/Edison1
If Arthur2/Edison2/Tintint2 is really dead, log on to Arthur1/Edison1/Tintin2 as root and run:
On Arthur1
# fake 168.156.85.16
On Edison1
# fake 168.156.85.15
On Tintin1
# fake 168.156.85.21
Switching from Arthur1/Edison1/Tintin1 back to Arthur2/Edison2/Tintin2
Log on to Arthur1/Edison1/Tintin1 as root.
On Arthur1
# fake remove 168.156.85.16
On Edison1
# fake remove 168.156.85.15
On Tintin1
# fake remove 168.156.85.21