Active/Active HA configuration

In an active/active configuration, one node is active running one or more ResourceGroup[1]s, and the other node is also running one or more ResourceGroup[1]s.

The disadvantage of this configuration is that quality of service may be degraded when one machine runs both services. Additionally, resource[2]s are moved not only when machines fail, but when they return to service, resulting in twice as many service transitions as active/passive configurations.

The advantage of this configuration is that both node[3]s are providing useful service, and is generally viewed as a more efficient use of resources than an active/passive[4] configuration.

To instruct Heartbeat[5] to operate in active/active mode for Release 1 (non-CRM[6]) clusters, simply make sure you configure the auto_failback[7] directive in your ha.cf[8] file like this:

auto_failback on

This will ensure that both machines will run services according to the preferred host[9] information in the haresources[10] file.

Typically each ResourceGroup[1] in the haresources[10] file will need to include at least one ServiceAddress[11] in it.

In release 2[12], simply create two groups, and Heartbeat[5] will naturally schedule them in an active/active arrangement. The auto_failback directive only applies to R1-style haresources[10] configurations, and should not be applied in R2[12] (CRM[6]-based) configurations.

The corresponding attribute to put in the CIB[13] is the resource_stickiness attribute.

See Also

Active/Passive[4] HA configuration

---

This information provided courtesy of the Linux-HA project at http://linux-ha.org/