Database high availability solution options: DRBD and Pacemaker

This article is from DBAnotes. Original article title: DRBD and Pacemaker

If someone asks you if a PC server can reach 99.99% high availability, what should you do? Maybe no machine can "Make sure" to achieve this availability. Of course, there may be no problems in a certain period of time, but it must be luck, however, high availability is basically impossible to achieve the goal through one machine. More often, we design solutions to ensure that the faulty machine is taken over as soon as possible when a problem occurs. Of course, this requires a greater cost.

For Oracle high Availability solutions, refer to the Maximum Availability Architecture (PAA) White Paper. However, Oracle does not recommend operating system-level solutions. The MySQL guiding policy is more flexible, and DRBD (Distributed Replicated Block Device) is an option to consider. I have been concerned about this before, but I know that there are not many cases in China. I don't know if it is in consideration of the speed limit on the NIC synchronization. Now there is a new transfer, and InfiniBand can be supported in version 8.3. After all, the original method of synchronizing data blocks through the NIC is limited by the NIC latency and bandwidth. The implementation of InfiniBand support is believed to win the trust of some enterprise users.

On Linux Kernel Summit 2009, I introduced DRBD this time (pay attention to the introduction of data consistency). Does this mean that we can officially enter the Kernel?

Compared with proprietary cluster management tools, Pacemaker, an open-source cluster management tool, supports Heartbeat and OpenAIS standards. Pacemaker can achieve more flexible Master/Slave, N + 1, N-N and other modes. The touching feeling is quite vital.

A good open-source solution is the design of active wooden houses, which is cheap, flexible, and environmentally friendly. Of course, it must be the first goal.